Andrew Burgess [Thu, 10 Feb 2022 09:59:23 +0000 (09:59 +0000)]
gdb/infrun: some extra infrun debug print statements
While reviewing a different patch I wanted to know more about what was
going on during GDB's stepping. I added some extra infrun debug print
calls, and I thought these might be useful to others.
Alan Modra [Wed, 9 Feb 2022 05:51:02 +0000 (16:21 +1030)]
Work around gcc-4 warnings in elf64-ppc.c
elf64-ppc.c: In function 'ppc64_elf_size_dynamic_sections':
elf64-ppc.c:10309:45: error: value computed is not used [-Werror=unused-value]
++lgot_ents, ++lgot_masks, isym != NULL && isym++)
It is of course a silly warning, fixed in later versions of gcc. I
wrote "isym != NULL && isym++" rather than the simpler "isym++" to
stop sanitisers complaining about incrementing a NULL pointer. isym
is of course unused in any code path where it might start off as
NULL. Sometimes you can't win. So don't try to be clever in reading
local symbols only when needed. 99 times out of 100 they will be
cached anyway.
* elf64-ppc.c (ppc64_elf_size_dynamic_sections): Avoid annoying
warnings by always reading local syms.
(ppc64_elf_layout_multitoc): Likewise.
Palmer Dabbelt [Mon, 7 Feb 2022 20:14:30 +0000 (12:14 -0800)]
RISC-V: Stop reporting warnings for mismatched extension versions
The extension version checking logic is really just too complicated to
encode into the linker, trying to do so causes more harm than good.
This removes the checks and the associated tests, leaving the logic to
keep the largest version of each extension linked into the target.
bfd/
* elfnn-riscv.c (riscv_version_mismatch): Rename to
riscv_update_subset_version, and stop reporting warnings on
version mismatches.
(riscv_merge_std_ext): Adjust calls to riscv_version_mismatch.
(riscv_merge_multi_letter_ext): Likewise.
Alan Modra [Tue, 8 Feb 2022 09:51:01 +0000 (20:21 +1030)]
PR28862, heap-buffer-overflow in parse_stab_string
I have no info on the format of a "SUNPRO C++ Namespace" stab, so am
relying on the previous code being correct in parsing these stabs.
Just don't allow NULs anywhere in the stab.
Andrew Burgess [Mon, 7 Feb 2022 17:13:34 +0000 (17:13 +0000)]
gdb: remove tailing newlines from index_cache_debug calls
I noticed that most of the calls to index_cache_debug include a
trailing newline. As the new debug mechanism already adds a newline,
that means all of these debug calls result in a blank line being
printed, which I think is a mistake.
Remove all the trailing newlines.
I also reformatted one of the index_cache_debug where a string will
now fit onto a single line.
Unless 'set debug index-cache on' is used, there should be no visible
change in output after this commit.
Andrew Burgess [Mon, 24 Jan 2022 15:29:49 +0000 (15:29 +0000)]
gdb/python: allow Value.format_string to return styled output
Add a new argument to the gdb.Value.format_string method, 'styling'.
This argument is False by default.
When this argument is True, then the returned string can contain output
styling escape sequences.
When this argument is False, then the returned string will not contain
any styling escape sequences.
If the returned string is going to be printed to the user, then it is
often nice to retain the GDB styling.
For the testing, we need to adjust the TERM environment variable, as
we do for all the styling tests. I'm now running all of the C tests
in gdb.python/py-format-string.exp in an environment where styling
could be generated, but only my new test should actually produce
styled output, hopefully this will catch the case where a bug might
cause format_string to always produce styled output.
Lancelot SIX [Tue, 11 Jan 2022 15:10:11 +0000 (10:10 -0500)]
gdb: make thread_info::m_thread_fsm a std::unique_ptr
While working on function calls, I realized that the thread_fsm member
of struct thread_info is a raw pointer to a resource it owns. This
commit changes the type of the thread_fsm member to a std::unique_ptr in
order to signify this ownership relationship and slightly ease resource
management (no need to manually call delete).
To ensure consistent use, the field is made a private member
(m_thread_fsm). The setter method (set_thread_fsm) can then check
that it is incorrect to associate a FSM to a thread_info object if
another one is already in place. This is ensured by an assertion.
The function run_inferior_call takes an argument as a pointer to a
call_thread_fsm and installs it in it in a thread_info instance. Also
change this function's signature to accept a unique_ptr in order to
signify that the ownership of the call_thread_fsm is transferred during
the call.
No user visible change expected after this commit.
Tested on x86_64-linux with no regression observed.
Andrew Burgess [Wed, 22 Dec 2021 12:57:44 +0000 (12:57 +0000)]
gdb: unbuffer all input streams when not using readline
This commit should fix PR gdb/28711. What's actually going on is
pretty involved, and there's still a bit of the story that I don't
understand completely, however, from my observed results, I think that
the change I propose making here (or something very similar) is going
to be needed.
The original bug report involves using eclipse to drive gdb using mi
commands. A separate tty is spun off in which to send gdb the mi
commands, this tty is created using the new-ui command.
The behaviour observed is that, given a particular set of mi commands
being sent to gdb, we sometimes see an ESPIPE error from a lseek
call, which ultimately results in gdb terminating.
The problems all originate from gdb_readline_no_editing_callback in
gdb/event-top.c, where we can (sometimes) perform calls to fgetc, and
allow glibc to perform buffering on the FILE object being used.
I say sometime, because, gdb_readline_no_editing_callback already
includes a call to disable the glibc buffering, but this is only done
if the input stream is not a tty. In our case the input stream is a
tty, so the buffering is left in place.
The first step to understanding why this problem occurs is to
understand that eclipse sends multiple commands to gdb very quickly
without waiting for and answer to each command, eclipse plans to
collect all of the command results after sending all the commands to
gdb. In fact, eclipse sends the commands to gdb that they appear to
arrive in the gdb process as a single block of data. When reproducing
this issue within the testsuite I find it necessary to send multiple
commands using a single write call.
The next bit of the story gets a little involved, and this is where my
understanding is not complete. I can describe the behaviour that I
observe, and (for me at least) I'm happy that what I'm seeing, if a
little strange, is consistent. In order to fully understand what's
going on I think I would likely need to dive into kernel code, which
currently seems unnecessary given that I'm happy with the solution I'm
proposing.
The following description all relates to input from a tty in which I'm
not using readline. I see the same problems either when using a
new-ui tty, or with gdb's standard, non-readline, mi tty.
Here's what I observe happening when I send multiple commands to gdb
using a single write, if I send gdb this:
command_1\ncommand_2\ncommand_3
then gdb's event loop will wake up (from its select) as it sees there
is input available. We call into gdb_readline_no_editing_callback,
where we call fgetc, glibc will do a single big read, and get back
just:
command_1\n
that is, despite there being multiple lines of input available, I
consistently get just a single line. From glibc a single character is
returned from the fgetc call, and within gdb we accumulate characters,
one at a time, into our own buffer. Eventually gdb sees the '\n'
character, and dispatches the whole 'command_1' into gdb's command
handler, which processes the command and prints the result. We then
return to gdb_readline_no_editing_callback, which in turn returns to
gdb's event loop where we re-enter the select.
Inside the select we immediately see that there is more input waiting
on the input stream, drop out of the select, and call back into
gdb_readline_no_editing_callback. In this function we again call
fgetc where glibc performs another big read. This time glibc gets:
command_2\n
that is, we once again get just a single line, despite there being a
third line available. Just like the first command we copy the whole
string, character by character into gdb's buffer, then handle the
command. After handling the command we go to the event loop, enter,
and then exit the select, and call back to the function
gdb_readline_no_editing_callback.
In gdb_readline_no_editing_callback we again call fgetc, this time
glibc gets the string:
command_3\n
like before, we copy this to gdb's buffer and handle the command, then
we return to the event loop. At this point the select blocks while we
wait for more input to arrive.
The important bit of this is that someone, somewhere is, it appears,
taking care to split the incoming write into lines.
My next experiment is to try something like:
this_is_a_very_long_command\nshort_command\n
However, I actually make 'this_is_a_very_long_command' very long, as
in many hundreds of characters long. One way to do this is:
echo xxxxxx.....xxxxx
and just adding more and more 'x' characters as needed. What I'm
aiming for is to have the first command be longer than glibc's
internal read buffer, which, on my machine, is 1024 characters.
However, for this discussion, lets imagine that glibc's buffer is just
8 characters (we can create just this situation by adding a suitable
setbuf call into gdb_readline_no_editing_callback).
Now, if I send gdb this data:
abcdefghij\nkl\n
The first read from glibc will get 'abcdefgh', that is a full 8
character buffer. Once gdb has copied these to its buffer we call
fgetc again, and now glibc will get 'ij\n', that is, just like before,
multiple lines are split at the '\n' character. The full command,
which is now in gdb's buffer can be handled 'abcdefghij', after which
we go (via the event loop) back to gdb_readline_no_editing_callback.
Now we call fgetc, and glibc will get 'kl\n', which is then handled in
the normal way.
So far, so good. However, there is, apparently, one edge case where
the above rules don't apply.
If the '\n' character is the first character read from the kernel,
then the incoming lines are not split up. So, given glibc's 8
character buffer, if I send gdb this:
abcdefgh\nkl\n
that is the first command is 8 characters plus a newline, then, on the
first read (from within glibc) we get 'abcdefgh' in a single buffer.
As there's no newline gdb calls fgetc again, and glibc does another
large read, now we get:
\nkl\n
which doesn't follow the above pattern - the lines are not split into
separate buffers!
So, gdb reads the first character from glibc using fgetc, this is the
newline. Now gdb has a complete command, and so the command is
handled. We then return to the event loop and enter the select.
The problem is that, as far as the kernel is concerned, there is no
more input pending, it's all been read into glibc's buffer, and so the
select doesn't return. The second command is basically stuck in
glibc's buffer.
If I send another command to gdb, or even just send an empty
command (a lone newline) then the select returns, we call into
gdb_readline_no_editing_callback, and now gdb sees the second
command.
OK, so the above is interesting, but it doesn't explain the ESPIPE
error.
Well, that's a slightly different, but related issue. The ESPIPE
case will _only_ show up when using new-ui to create the separate tty
for mi commands, and is a consequence of this commit:
Prior to this commit, the new-ui command would open the tty three
times, once each for stdin, stderr, and stdout. After this commit we
open the tty just once and reuse the FILE object for all three roles.
Consider the problem case, where glibc has (unexpectedly) read the
second command into its internal buffer. When we handle the first
command we usually end up having to write something to the mi output
stream.
After the above commit the same FILE object represents both the input
and output streams, so, when gdb tries to write to the FILE object,
glibc spots that there is input pending within the input buffer, and
so assumes that we have read ahead of where we should be in the input
file. To correct for this glibc tries to do an lseek call to
reposition the file offset of the output stream prior to writing to
it. However, as the output stream is a tty, and seeking is not
supported on a tty, this lseek call fails, this results in the ESPIPE,
which ultimately causes gdb to terminate.
So, now we understand why the ESPIPE triggers (which was what caused
the gdb crash in the original bug report), and we also understand that
sometime gdb will not handle the second command in a timely
fashion (if the first command is just the wrong length). So, what to
do about all this?
We could revert the commit mentioned above (and implement its
functionality another way). This would certainly resolve the ESPIPE
issue, the buffered input would now only be on the input stream, the
output stream would have no buffered input, and so glibc would never
try to lseek, and so we'd never get the ESPIPE error.
However, this only solves one of the two problems. We would still
suffer from the problem where, if the first command is just the wrong
length, the second command will not (immediately) get handled.
The only solution I can see to this problem is to unbuffer the input
stream. If glibc is not buffering the input, but instead, we read
incoming data character by character from the kernel, then everything
will be fine. As soon as we see the newline at the end of the first
command we will handle the first command. As glibc will have no
buffered input it will not be tempted to lseek, so no ESPIPE error.
When we go have to the event loop there will be more data pending in
the kernel, so the select will immediately return, and the second
command will be processed.
I'm tempted to suggest that we should move the unbuffering of the
input stream out of gdb_readline_no_editing_callback and do it
somewhere earlier, more like when we create the input streams.
However, I've not done that in this commit for a couple of reasons:
1. By keeping the unbuffering in gdb_readline_no_editing_callback
I'm making the smallest possible change that fixes the bug. Moving
the unbuffering somewhere better can be done as a refactor later, if
that 's felt to be important,
2. I don't think making repeated calls to unbuffer the input will
have that much performance impact. We only make the unbuffer call
once per call to gdb_readline_no_editing_callback, and, if the input
stream is already unbuffered we'll return pretty quickly, so I don't
see this as being massively costly,
3. Tom is currently doing lots of gdb stream management changes and
I want to minimise the chances we'll conflict.
So, this commit just changes gdb_readline_no_editing_callback to
always unbuffer the input stream.
The test for this issue sends two commands in a loop, with the first
command growing bigger each time around the loop. I actually make the
first command bigger by just adding whitespace to the front, as gdb
still has to read the complete command (including whitespace) via
glibc, so this is enough to trigger the bug.
The original bug was reported when using a virtual machine, and in
this situation we see this in the strace output:
I'm not completely sure what's going on here, but it appears that the
kernel on the virtual machine is delivering the input to glibc slower
than I see on my real hardware; glibc asks for 1024 bytes, but only
gets 64 bytes the first time. In the second read we see the problem
case, the first character is the newline, but then the entire second
command is included.
If I run this exact example on my real hardware then the first command
would not be truncated at 64 bytes, instead, I'd expect to see the
newline included in the first read, with the second command split into
a second read.
So, for testing, I check cases where the first command is just a few
characters (starting at 8 character), all the way up to 2048
characters. Hopefully, this should mean we hit the problem case for
most machine setups.
The only last question relates to commit afe09f0b6311a4d that I
mentioned earlier. That commit was intended to provide support for
Microsoft named pipes:
I know next to nothing about this topic beyond a brief scan of the
above link, but I think these windows named pipe are closer in
behaviour to unix sockets than to unix named fifos.
I am a little nervous that, after the above commit, we now use the
same FILE for in, err, and out streams. In contrast, in a vanilla C
program, I would expect different FILE objects for each stream.
Still, I'm reluctant to revert the above commit (and provide the same
functionality a different way) without a specific bug to point at,
and, now that the streams are unbuffered, I expect a lot of the read
and write calls are going straight to the kernel with minimal glibc
involvement, so maybe it doesn't really matter. Anyway, I haven't
touched the above patch, but it is something to keep in mind when
working in this area.
Andrew Burgess [Fri, 4 Feb 2022 16:51:45 +0000 (16:51 +0000)]
gdb/disasm: combine the no printing disassembler setup code
We have three places in gdb where we initialise a disassembler that
will not print anything (used for figuring out the length of
instructions, or collecting other information from the disassembler).
Each of these places has its own stub function to act as a print like
callback, the stub function is identical in each case, and just does
nothing.
In this commit I create a new function to initialise a disassembler
that doesn't print anything, and have all three locations use this new
function. There's now only one non-printing stub function.
There should be no user visible changes after this commit.
gdb: add the 'set/show suppress-cli-notifications' command
GDB already has a flag to suppress printing notification events, such
as thread and inferior context switches, on the CLI. This is used
internally when executing commands. Make the flag available to the
user via a new command. This is expected to be useful in scripts.
For instance, suppose that when Inferior 1 gets to a certain state,
you want to add and set up a new inferior using the commands below,
but you also want to have a reduced/clean output.
define do-setup
printf "Setting up Inferior 2...\n"
add-inferior -exec a.out
inferior 2
break file.c:3
run
inferior 1
printf "Done\n"
end
Currently, GDB prints
(gdb) do-setup
Setting up Inferior 2...
[New inferior 2]
Added inferior 2 on connection 1 (native)
[Switching to inferior 2 [<null>] (/tmp/a.out)]
Breakpoint 2 at 0x1155: file file.c, line 3.
Thread 2.1 "a.out" hit Breakpoint 2, main () at file.c:3
3 return 0;
[Switching to inferior 1 [process 7670] (/tmp/test)]
[Switching to thread 1.1 (process 7670)]
#0 main () at test.c:2
2 int a = 1;
Done
GDB's Python API make it possible to capture and return GDB's output,
but this does not work for all the streams. In particular, CLI
notification events are not captured:
Thread 2.1 "a.out" hit Breakpoint 2, main () at file.c:3
3 return 0;
[Switching to inferior 1 [process 8263] (/tmp/test)]
[Switching to thread 1.1 (process 8263)]
#0 main () at test.c:2
2 int a = 1;
You can use the new "set suppress-cli-notifications" command to
suppress the output:
(gdb) set suppress-cli-notifications on
(gdb) do-setup
Setting up Inferior 2...
[New inferior 2]
Added inferior 2 on connection 1 (native)
Breakpoint 2 at 0x1155: file file.c, line 3.
Done
gdb/cli: add a 'normal_stop' option to 'cli_suppress_notification'
Extend the 'cli_suppress_notification' struct with a new field,
'normal_stop', that can be used for checking if printing normal stop
events on the CLI should be suppressed.
This patch only introduces the flag. The subsequent patch adds a user
command to turn the flag off/on.
Tom Tromey [Thu, 27 Jan 2022 00:57:33 +0000 (17:57 -0700)]
Merge do_val_print and common_val_print
The only caller of do_val_print just does a small bit of work before
the call. This patch merges the two functions, and removes an
unnecessary local variable, making gdb a bit simpler.
Add a getter and a setter for a symtab's compunit_symtab. Remove the
corresponding macro and adjust all callers.
For brevity, I chose the name "compunit" instead of "compunit_symtab"
the the field, getter and setter names. Since we are already in symtab
context, the _symtab suffix seems redundant.
Simon Marchi [Sat, 20 Nov 2021 02:35:17 +0000 (21:35 -0500)]
gdb: remove COMPUNIT_FILETABS macro
I think that most remaining uses of COMPUNIT_FILETABS intend to get the
primary filetab of the compunit_symtab specifically (and not to iterate
over all filetabs, for example, those cases would use compunit_filetabs,
which has been converted to compunit_symtab::filetabs), so replace mosts
uses with compunit_symtab::primary_filetab.
In jit.c, function finalize_symtab, we can save the symtab object
returned by allocate_symtab and use it, it makes things simpler.
Simon Marchi [Sat, 20 Nov 2021 02:18:05 +0000 (21:18 -0500)]
gdb: move compunit_filetabs to compunit_symtab::filetabs
Make compunit_filetabs, used to iterate a compunit_symtab's filetabs, a
method of compunit_symtab. The name filetabs conflicts with the current
name of the field. Rename the field to m_filetabs, since at this point
nothing outside of compunit_symtab uses it, so we should treat it as
private (even though it's not actually private). Rename the
last_filetab field to m_last_filetab as well (it's only used on
compunit_symtab::add_filetab).
Adjust the COMPUNIT_FILETABS macro to keep its current behavior of
returning the first filetab.
Tom Tromey [Wed, 26 Jan 2022 22:39:03 +0000 (15:39 -0700)]
Allow non-ASCII characters in Rust identifiers
Rust 1.53 (quite a while ago now) ungated the support for non-ASCII
identifiers. This didn't work in gdb. This is PR rust/20166.
This patch fixes the problem by allowing non-ASCII characters to be
considered as identifier components. It seemed simplest to just pass
them through -- doing any extra checking didn't seem worthwhile.
The new test also verifies that such characters are allowed in strings
and character literals as well. The latter also required a bit of
work in the lexer.
Tom Tromey [Tue, 25 Jan 2022 01:02:38 +0000 (18:02 -0700)]
Fix Rust parser bug with function fields
In Rust, 'obj.f()' is a method call -- but '(obj.f)()' is a call of a
function-valued field 'f' in 'obj'. The Rust parser in gdb currently
gets this wrong. This is PR rust/24082.
The expression and Rust parser rewrites made this simple to fix --
simply wrapping a parenthesized expression in a new operation handles
it. This patch has a slight hack because I didn't want to introduce a
new exp_opcode enumeration constant just for this. IMO this doesn't
matter, since we should work toward removing dependencies on these
opcodes anyway; but let me know what you think of this.
Andrew Burgess [Fri, 28 Jan 2022 11:49:54 +0000 (11:49 +0000)]
gdb/doc: update docs for 'info win' and 'winheight' commands
This started by noticing that the docs for 'winheight' are out of
date, the docs currently give a specific list of possible window
names. However, now that windows can be implemented in Python, it is
not possible to list all possible names.
I now link the user to a mechanism by which they can discover the
valid names for themselves at run time (by using 'info win'). That,
and the fact that gdb provides tab-completion of the name at the
command line, feels good enough.
Finally, I noticed that the docs for 'win info' don't explicitly say
that the name of the window is given in the output. This could
probably have been inferred, but given I'm now linking to this as a
mechanism to find the window name, I'd prefer to mention that the name
can be found in the output.
Andrew Burgess [Mon, 24 Jan 2022 21:25:38 +0000 (21:25 +0000)]
gdb/tui: add window width information to 'info win' output
Now that we support horizontal window placement in the tui, it makes
sense to have 'info win' include the width, as well as the height, of
the currently visible windows.
That's what this commit does. Example output is now:
(gdb) info win
Name Lines Columns Focus
src 12 40 (has focus)
asm 12 41
status 1 80
cmd 11 80
I've added a NEWS entry, but the documentation was already suitably
vague, it just says that 'info win' displays the size of the visible
windows, so I don't think anything needs to be added there.
I've also added some tests, as far as I could find, the 'info win'
command was previously untested.
Alan Modra [Fri, 4 Feb 2022 09:00:47 +0000 (19:30 +1030)]
Detect .eh_frame_hdr earlier for SIZEOF_HEADERS
Current code detects the need for PT_GNU_EH_FRAME using a field set by
_bfd_elf_discard_section_eh_frame_hdr, which is called fairly late in
the linking process. Use the elf hash table eh_info instead, which is
set up earlier by size_dynamic_sections.
sim/testsuite/mips/ChangeLog:
* basic.exp: Add r6-*.s tests.
(run_r6_removed_test): New function.
(run_endian_tests): New function.
* hilo-hazard-3.s: Skip for mips*r6.
* r2-fpu.s: New test.
* r6-64.s: New test.
* r6-branch.s: New test.
* r6-forbidden.s: New test.
* r6-fpu.s: New test.
* r6-llsc-dp.s: New test.
* r6-llsc-wp.s: New test.
* r6-removed.csv: New test.
* r6-removed.s: New test.
* r6.s: New test.
* utils-r6.inc: New inc.
sim/common/ChangeLog:
* sim-fpu.c (sim_fpu_op_nan): New.
(sim_fpu_add): Factor out NaN operand handling with
a call to sim_fpu_op_nan.
(sim_fpu_sub, sim_fpu_mul, sim_fpu_div): Likewise.
(sim_fpu_rem, sim_fpu_max, sim_fpu_min): Likewise.
* sim-fpu.h (sim_fpu_op_nan): New declaration.
IEEE754-1985 specifies the top bit of the mantissa as an indicator
of signalling vs. quiet NaN, but does not define the precise semantics.
Most architectures treat this bit as indicating quiet NaN, but legacy
(pre-R6) MIPS goes the other way and treats it as signalling NaN.
This used to be controlled by a macro that was only defined for MIPS.
This patch replaces the macro with a variable to track the current
semantics of the NaN bit and allows differentiation between older
(pre-R6) and and newer MIPS cores.
sim/common/ChangeLog:
* sim-fpu.c (_sim_fpu): New.
(pack_fpu, unpack_fpu): Allow reversal of quiet NaN semantics.
* sim-fpu.h (sim_fpu_state): New struct.
(_sim_fpu): New extern.
(sim_fpu_quiet_nan_inverted): New define.
This also removes arm-netbsd (not arm-netbsdelf!), arm-openbsd, and
arm-riscix. Those targets weren't on the obsolete list but they are
all aout, and it doesn't make all that much sense to remove arm-aout
without removing them too.
Simon Marchi [Wed, 2 Feb 2022 15:54:03 +0000 (10:54 -0500)]
gdb: include jit_code_entry::symfile_addr value in names of objfiles created by jit reader API
This commit includes the JIT object's symfile address in the names of
objfiles created by JIT reader API (e.g., << JIT compiled code at
0x7ffd8a0c77a0 >>). This allows one to at least differentiate one from
another.
The address is the one that the debugged program has put in
jit_code_entry::symfile_addr, and that the JIT reader's read function
receives. As we can see in gdb.base/jit-reader-host.c and
gdb.base/jit-reader.c, that may not be the actual value of where the
JIT-ed code is. But it is a value chosen by the author of the JIT
engine and the JIT reader, so including this value in the objfile name
may help them correlate the JIT objfiles created by with their logs /
data structures.
To access this field, we need to pass down a reference to the
jit_code_entry. So make jit_dbg_reader_data a structure (instead of an
alias for a CORE_ADDR) that includes the address of the code entry in
the inferior's address space (the previous meaning of
jit_dbg_reader_data) plus a reference to the jit_code_entry as read into
GDB's address space. And while at it, pass down the gdbarch, so that we
don't have to call target_gdbarch.
Co-Authored-By: Jan Vrany <jan.vrany@labware.com>
Change-Id: Ib26c4d1bd8de503d651aff89ad2e500cb312afa5
Tom Tromey [Tue, 1 Feb 2022 16:46:43 +0000 (09:46 -0700)]
Improve Ada unchecked union type printing
Currently, "ptype" of an Ada unchecked union may show a
compiler-generated wrapper structure in its output. It's more
Ada-like to elide this structure, which is what this patch implements.
It turned out to be simplest to reuse a part of print_variant_clauses
for this.
As this is Ada-specific, and Joel already reviewed it internally, I am
going to check it in.
Andi Kleen [Thu, 23 Dec 2021 17:55:07 +0000 (09:55 -0800)]
Support symbol+offset lookup in addr2line
The Linux kernel usually ouputs symbol+offset instead of plain code
addresses these days, to avoid leaking ASLR secrets and to handle
dynamically loaded modules.
Converting those with addr2line is somewhat involved: it requires
looking up the symbol first using nm and then manually compute the
offset, and then pass it to addr2line.
This patch implements the necessary steps directly in addr2line,
by looking up the symbol (with demangling if needed) and computing
the offset.
It's possible that a symbol is ambigious with a hex number. In this
case it uses the symbol lookup if the string contains a +. When it isn't
ambigious the + is optional.
Change how Python architecture and language are handled
we see a failure in gdb.threads/killed-outside.exp:
...
Executing on target: kill -9 16622 (timeout = 300)
builtin_spawn -ignore SIGHUP kill -9 16622
continue
Continuing.
Couldn't get registers: No such process.
(gdb) [Thread 0x7ffff77c2700 (LWP 16626) exited]
Program terminated with signal SIGKILL, Killed.
The program no longer exists.
FAIL: gdb.threads/killed-outside.exp: prompt after first continue (timeout)
This is not a regression but a failure due to a change in GDB's
output. Prior to the aforementioned commit, GDB has been printing the
"Couldn't get registers: No such process." message twice. The second
one came from
(top-gdb) bt
#0 amd64_linux_nat_target::fetch_registers (this=0x555557f31440 <the_amd64_linux_nat_target>, regcache=0x555558805ce0, regnum=16) at /gdb-up/gdb/amd64-linux-nat.c:225
#1 0x000055555640ac5f in target_ops::fetch_registers (this=0x555557d636d0 <the_thread_db_target>, arg0=0x555558805ce0, arg1=16) at /gdb-up/gdb/target-delegates.c:502
#2 0x000055555641a647 in target_fetch_registers (regcache=0x555558805ce0, regno=16) at /gdb-up/gdb/target.c:3945
#3 0x0000555556278e68 in regcache::raw_update (this=0x555558805ce0, regnum=16) at /gdb-up/gdb/regcache.c:587
#4 0x0000555556278f14 in readable_regcache::raw_read (this=0x555558805ce0, regnum=16, buf=0x555558881950 "") at /gdb-up/gdb/regcache.c:601
#5 0x00005555562792aa in readable_regcache::cooked_read (this=0x555558805ce0, regnum=16, buf=0x555558881950 "") at /gdb-up/gdb/regcache.c:690
#6 0x000055555627965e in readable_regcache::cooked_read_value (this=0x555558805ce0, regnum=16) at /gdb-up/gdb/regcache.c:748
#7 0x0000555556352a37 in sentinel_frame_prev_register (this_frame=0x555558181090, this_prologue_cache=0x5555581810a8, regnum=16) at /gdb-up/gdb/sentinel-frame.c:53
#8 0x0000555555fa4773 in frame_unwind_register_value (next_frame=0x555558181090, regnum=16) at /gdb-up/gdb/frame.c:1235
#9 0x0000555555fa420d in frame_register_unwind (next_frame=0x555558181090, regnum=16, optimizedp=0x7fffffffd570, unavailablep=0x7fffffffd574, lvalp=0x7fffffffd57c, addrp=0x7fffffffd580,
realnump=0x7fffffffd578, bufferp=0x7fffffffd5b0 "") at /gdb-up/gdb/frame.c:1143
#10 0x0000555555fa455f in frame_unwind_register (next_frame=0x555558181090, regnum=16, buf=0x7fffffffd5b0 "") at /gdb-up/gdb/frame.c:1199
#11 0x00005555560178e2 in i386_unwind_pc (gdbarch=0x5555587c4a70, next_frame=0x555558181090) at /gdb-up/gdb/i386-tdep.c:1972
#12 0x0000555555cd2b9d in gdbarch_unwind_pc (gdbarch=0x5555587c4a70, next_frame=0x555558181090) at /gdb-up/gdb/gdbarch.c:3007
#13 0x0000555555fa3a5b in frame_unwind_pc (this_frame=0x555558181090) at /gdb-up/gdb/frame.c:948
#14 0x0000555555fa7621 in get_frame_pc (frame=0x555558181160) at /gdb-up/gdb/frame.c:2572
#15 0x0000555555fa7706 in get_frame_address_in_block (this_frame=0x555558181160) at /gdb-up/gdb/frame.c:2602
#16 0x0000555555fa77d0 in get_frame_address_in_block_if_available (this_frame=0x555558181160, pc=0x7fffffffd708) at /gdb-up/gdb/frame.c:2665
#17 0x0000555555fa5f8d in select_frame (fi=0x555558181160) at /gdb-up/gdb/frame.c:1890
#18 0x0000555555fa5bab in lookup_selected_frame (a_frame_id=..., frame_level=-1) at /gdb-up/gdb/frame.c:1720
#19 0x0000555555fa5e47 in get_selected_frame (message=0x0) at /gdb-up/gdb/frame.c:1810
#20 0x0000555555cc9c6e in get_current_arch () at /gdb-up/gdb/arch-utils.c:848
#21 0x000055555625b239 in gdbpy_before_prompt_hook (extlang=0x555557451f20 <extension_language_python>, current_gdb_prompt=0x555557f4d890 <top_prompt+16> "(gdb) ")
at /gdb-up/gdb/python/python.c:1063
#22 0x0000555555f7cfbb in ext_lang_before_prompt (current_gdb_prompt=0x555557f4d890 <top_prompt+16> "(gdb) ") at /gdb-up/gdb/extension.c:922
#23 0x0000555555f7d442 in std::_Function_handler<void (char const*), void (*)(char const*)>::_M_invoke(std::_Any_data const&, char const*&&) (__functor=...,
__args#0=@0x7fffffffd900: 0x555557f4d890 <top_prompt+16> "(gdb) ") at /usr/include/c++/7/bits/std_function.h:316
#24 0x0000555555f752dd in std::function<void (char const*)>::operator()(char const*) const (this=0x55555817d838, __args#0=0x555557f4d890 <top_prompt+16> "(gdb) ")
at /usr/include/c++/7/bits/std_function.h:706
#25 0x0000555555f75100 in gdb::observers::observable<char const*>::notify (this=0x555557f49060 <gdb::observers::before_prompt>, args#0=0x555557f4d890 <top_prompt+16> "(gdb) ")
at /gdb-up/gdb/../gdbsupport/observable.h:150
#26 0x0000555555f736dc in top_level_prompt () at /gdb-up/gdb/event-top.c:444
#27 0x0000555555f735ba in display_gdb_prompt (new_prompt=0x0) at /gdb-up/gdb/event-top.c:411
#28 0x00005555564611a7 in tui_on_command_error () at /gdb-up/gdb/tui/tui-interp.c:205
#29 0x0000555555c2173f in std::_Function_handler<void (), void (*)()>::_M_invoke(std::_Any_data const&) (__functor=...) at /usr/include/c++/7/bits/std_function.h:316
#30 0x0000555555e10c20 in std::function<void ()>::operator()() const (this=0x5555580f9028) at /usr/include/c++/7/bits/std_function.h:706
#31 0x0000555555e10973 in gdb::observers::observable<>::notify() const (this=0x555557f48d20 <gdb::observers::command_error>) at /gdb-up/gdb/../gdbsupport/observable.h:150
#32 0x00005555560e9b3f in start_event_loop () at /gdb-up/gdb/main.c:438
#33 0x00005555560e9bcc in captured_command_loop () at /gdb-up/gdb/main.c:481
#34 0x00005555560eb616 in captured_main (data=0x7fffffffddd0) at /gdb-up/gdb/main.c:1348
#35 0x00005555560eb67c in gdb_main (args=0x7fffffffddd0) at /gdb-up/gdb/main.c:1363
#36 0x0000555555c1b6b3 in main (argc=12, argv=0x7fffffffded8) at /gdb-up/gdb/gdb.c:32
Commit 1da5d0e664 eliminated the call to 'get_current_arch'
in 'gdbpy_before_prompt_hook'. Hence, the second instance of
"Couldn't get registers: No such process." does not appear anymore.
Fix the failure by updating the regular expression in the test.
Alan Modra [Wed, 26 Jan 2022 00:01:36 +0000 (10:31 +1030)]
PowerPC64 treatment of absolute symbols
Supporting -static-pie on PowerPC64 requires the linker to properly
treat SHN_ABS symbols for cases like glibc's _nl_current_LC_CTYPE_used
absolute symbol. I've been slow to fix the linker on powerpc because
there is some chance that this will break some shared libraries or
PIEs.
bfd/
* elf64-ppc.c (ppc64_elf_check_relocs): Consolidate local sym
handling code. Don't count dyn relocs against non-dynamic
absolute symbols.
(dec_dynrel_count): Adjust to suit.
(ppc64_elf_edit_toc): Don't remove entries for absolute symbols
when pic.
(allocate_got): Don't allocate space for got relocs against
non-dynamic absolute syms.
(ppc64_elf_layout_multitoc): Likewise.
(got_and_plt_relr): Likewise.
(ppc64_elf_size_dynamic_sections): Likewise for local got.
(got_and_plt_relr_for_local_syms): Likewise.
(ppc64_elf_size_stubs): Don't allocate space for relr either.
(ppc64_elf_relocate_section): Don't write relocs against non-dynamic
absolute symbols. Don't optimise got and toc code sequences
loading absolute symbol entries.
ld/
* testsuite/ld-powerpc/abs-reloc.s,
* testsuite/ld-powerpc/abs-static.d,
* testsuite/ld-powerpc/abs-static.r,
* testsuite/ld-powerpc/abs-pie.d,
* testsuite/ld-powerpc/abs-pie.r,
* testsuite/ld-powerpc/abs-shared.d,
* testsuite/ld-powerpc/abs-shared.r,
* testsuite/ld-powerpc/abs-pie-relr.d,
* testsuite/ld-powerpc/abs-pie-relr.r,
* testsuite/ld-powerpc/abs-shared-relr.d,
* testsuite/ld-powerpc/abs-shared-relr.r: New tests.
* testsuite/ld-powerpc/powerpc.exp: Run them.
Andrew Burgess [Tue, 1 Feb 2022 21:46:29 +0000 (21:46 +0000)]
gdb: fix formatting for help set/show extended-prompt
The formatting of the help text for 'help set extended-prompt' and
'help show extended-prompt' is a little off.
Here's the offending snippet:
Substitutions are applied to VALUE to compute the real prompt.
The currently defined substitutions are:
\[ Begins a sequence of non-printing characters.
\\ A backslash.
\] Ends a sequence of non-printing characters.
\e The ESC character.
Notice that the line for '\[' is indented more that the others.
Turns out this is due to how we build this help text, something which
is done in Python. We extended a classes __doc__ string with some
dynamically generated text.
The classes doc string looks like this:
"""Set the extended prompt.
Usage: set extended-prompt VALUE
Substitutions are applied to VALUE to compute the real prompt.
The currently defined substitutions are:
"""
Notice the closing """ are in a line of their own, and include some
white space just before. It's this extra white space that's causing
the problem.
Fix the formatting issue by moving the """ to the end of the previous
line. I then add the extra newline in at the point where the doc
string is merged with the dynamically generated text.
Andrew Burgess [Thu, 27 Jan 2022 15:12:25 +0000 (15:12 +0000)]
gdb: test to check one aspect of the linespec parsing code
While working on the fix for PR cli/28665 (see previous couple of
commits), I was playing with making a change in the linespec parsing
code. Specifically, I was thinking about whether the spec_string for
LINESPEC_LOCATION locations should ever be nullptr.
I made a change to prevent the spec_string from ever being nullptr,
tested gdb, and saw no regressions.
However, as part of this work I was reviewing how the breakpoint code
handles this case (spec_string being nullptr), and spotted that in
parse_breakpoint_sals the nullptr case is specifically handled, so
changing this should have caused a regression. But I didn't see one.
So, this commit adds a comment in location.c mentioning that the
nullptr case is (a) not an oversight, and (b) is required. Then I add
a new test to gdb.base/break.exp that ensures a change in this area
will cause a regression.
This test passes on current gdb, but with my modified (and broken)
gdb, the test would fail.
Andrew Burgess [Tue, 7 Dec 2021 22:26:05 +0000 (22:26 +0000)]
gdb: handle calls to edit command passing only a linespec condition
While working on the previous commit to fix PR cli/28665, I noticed
that the 'edit' command would suffer from the same problem. That is,
something like:
(gdb) edit task 123
would cause GDB to break. For a full explanation of what's going on
here, see the commit message for the previous commit.
As with the previous commit, this issue can be prevented by detecting,
and throwing, a junk at the end of the line error earlier, before
calling decode_line_1.
So, that's what this commit does. I've also added some tests for this
issue.
Andrew Burgess [Tue, 7 Dec 2021 14:01:23 +0000 (14:01 +0000)]
gdb: handle calls to list command passing only a linespec condition
In PR cli/28665, it was reported that GDB would crash when given a
command like:
(gdb) list task 123
The problem here is that in cli/cli-cmd.c:list_command, the string
'task 123' is passed to string_to_event_location in find a location
specification. However, this location parsing understands about
breakpoint conditions, and so, will stop parsing when it sees
something that looks like a condition, in this case, the 'task 123'
looks like a breakpoint condition.
As a result, the location we get back from string_to_event_location
has no actual location specification attached to it. The actual call
path is:
In new_linespec_location we call linespec_lex_to_end, which looks at
'task 123' and decides that there's nothing there that describes a
location. As such, in new_linespec_location, the spec_string field of
the location is left as nullptr.
Back in list_command we then call decode_line_1, which calls
event_location_to_sals, which calls parse_linespec, which takes the
spec_string we found earlier, and tries to converts this into a list
of sals.
However, parse_linespec is not intended to be passed a nullptr, for
example, calling is_ada_operator will try to access through the
nullptr, causing undefined behaviour. But there are other cases
within parse_linespec which don't expect to see a nullptr.
When looking at how to fix this issue, I first considered having
linespec_lex_to_end detect the problem. That function understands
when the first thing in the linespec is a condition keyword, and so,
could throw an error saying something like: "no linespec before
condition keyword", however, this is not going to work, at least, not
without additional changes to GDB, it is valid to place a breakpoint
like:
(gdb) break task 123
This will place a breakpoint at the current location with the
condition 'task 123', and changing linespec_lex_to_end breaks this
behaviour.
So, next, I considered what would happen if I added a condition to an
otherwise valid list command, this is what I see:
(gdb) list file.c:1 task 123
Junk at end of line specification.
(gdb)
So, then I wondered, could we just pull the "Junk" detection forward,
so that we throw the error earlier, before we call decode_line_1?
It turns out that yes we can. Well, sort of.
It is simpler, I think, to add a separate check into the list_command
function, after calling string_to_event_location, but before calling
decode_line_1. We know when we call string_to_event_location that the
string in question is not empty, so, after calling
string_to_event_location, if non of the string has been consumed, then
the content of the string must be junk - it clearly doesn't look like
a location specification.
I've reused the same "Junk at end of line specification." error for
consistency, and added a few tests to cover this issue.
While the first version of this patch was on the mailing list, a
second bug PR gdb/28797 was raised. This was for a very similar
issue, but this time the problem command was:
(gdb) list ,,
Here the list command understands about the first comma, list can have
two arguments separated by a comma, and the first argument can be
missing. So we end up trying to parse the second command "," as a
linespec.
However, in linespec_lex_to_end, we will stop parsing a linespec at a
comma, so, in the above case we end up with an empty linespec (between
the two commas), and, like above, this results in the spec_string
being nullptr.
As with the previous case, I've resolved this issue by adding an extra
check for junk at the end of the line - after parsing (or failing to
parse) the nothing between the two commas, we still have the "," left
at the end of the list command line - when we see this we can throw
the same "junk at the end of the line" error, and all is good.
Andrew Burgess [Tue, 7 Dec 2021 22:31:11 +0000 (22:31 +0000)]
gdb/testsuite: move linespec test into gdb.linespec/ directory
The gdb.base/linespecs.exp test should really live in the gdb.linespec
directory, so lets move it there.
As we already have gdb.linespec/linespec.exp, I've renamed the test to
gdb.linespec/errors.exp, as this better reflects what the test is
actually checking.
Finally, the test script doesn't have its own source file, it was
reusing a random other source file, gdb.base/memattr.c. Now the tests
script is in gdb.linespec/, I've updated the test to use a different
source file from that directory.
Andrew Burgess [Tue, 7 Dec 2021 13:25:47 +0000 (13:25 +0000)]
gdb: add empty string check in parse_linespec
If parse_linespec (linespec.c) is passed ARG as an empty string then
we end up calling `strchr (linespec_quote_characters, '\0')`, which
will return a pointer to the '\0' at the end of
linespec_quote_characters. This then results in GDB calling
skip_quote_char with `ARG + 1`, which is undefined behaviour (as ARG
only contained a single character, the '\0').
Fix this by checking for the first character of ARG being '\0' before
the call to strchr.
I have additionally added an assertion that ARG can't itself be
nullptr, as calling is_ada_operator with nullptr can end up calling
'startswith' on the nullptr, which is undefined behaviour.
Finally, I moved the declaration of TOKEN into the body of
parse_linespec, to where TOKEN is defined.
This patch came about while I was working on fixes for PR cli/28665
and PR gdb/28797. The actual fixes for these two issues will be in a
later commit in this series, but, with this patch in place, both of
the above bugs would hit the new assertion rather than accessing
invalid memory and crashing. The '\0' check is not currently ever
hit, but just makes the code a little safer.
Because this patch only changes the nature of the failure for the
above two bugs, there's no tests here. A later commit will fix the
above two issues, at which point I'll add some tests.
Andrew Burgess [Tue, 7 Dec 2021 13:22:33 +0000 (13:22 +0000)]
gdb: update the comment on string_to_event_location
The comment on string_to_event_location is (I believe) out of date.
This commit fixes the two issues I see:
1. This function can't return NULL any more. The implementation
calls string_to_explicit_location which can return NULL, but if this
is the case we then call string_to_event_location_basic, which I
don't believe can ever return NULL.
2. I've removed the mention that the returned string is malloc'd,
though this is true, now that we return a managed pointer, I believe
the source of the memory allocation is irrelevant, and so, shouldn't
be discussed in the header comment.
There should be no user visible changes after this commit.
Stafford Horne [Wed, 2 Feb 2022 11:11:56 +0000 (20:11 +0900)]
or1k: Avoid R_OR1K_GOT16 signed overflow by using special howto
Previously when fixing PR 21464 we masked out upper bits of the
relocation value in order to avoid overflow complaints when acceptable.
It turns out this does not work when the relocation value ends up being
signed.
To fix this this patch introduces a special howto with
complain_on_overflow set to complain_overflow_dont. This is used in
place of the normal R_OR1K_GOT16 howto when we detect R_OR1K_GOT_AHI16
relocations.
bfd/ChangeLog:
PR 28735
* elf32-or1k.c (or1k_elf_got16_no_overflow_howto): Define.
(or1k_elf_relocate_section): Use new howto instead of trying to
mask out relocation bits.
Tom Tromey [Tue, 25 Jan 2022 21:57:18 +0000 (14:57 -0700)]
Fix flex rule in gdb
Currently, if flex fails, it will leave the resulting .c file in the
tree. This will cause a cascade of errors, and requires the manual
deletion of the .c file in order to recreate the problem.
It's better for the rule to fail such that the .c file is not updated.
This way, 'make' will fail the same way every time -- which is much
handier for debugging syntax errors.
This fix just updates the Makefile rule to follow the way that the
"yacc" rule already works.
Markus Metzger [Mon, 31 Jan 2022 06:24:54 +0000 (07:24 +0100)]
gdb, btrace: improve error messages
When trying to use 'record btrace' on a system that does not support it,
the error message isn't as clear as it could be. See
https://sourceware.org/pipermail/gdb/2022-January/049870.html.
Jan Vrany [Tue, 1 Feb 2022 14:49:30 +0000 (14:49 +0000)]
gdb/python: fix gdb.Objfile.__repr__ () for dynamically compiled code
While experimenting with JIT reader API I realized that calling repr ()
on objfile created by JIT reader crashes GDB.
The problem was that objfpy_repr () called objfile_filename () which
returned NULL, causing PyString_FromFormat () to crash.
This commit fixes this problem by using objfile_name () instead of
objfile_filename (). This also makes consistent with the value of gdb.Objfile.filename variable.
projects / thirdparty / binutils-gdb.git / log
