Tom Tromey [Fri, 23 Apr 2021 17:28:48 +0000 (11:28 -0600)]
Fix crash when expanding partial symtabs with DW_TAG_imported_unit
PR gdb/27743 points out a gdb crash when expanding partial symtabs,
where one of the compilation units uses DW_TAG_imported_unit.
The bug is that partial_map_expand_apply expects only to be called for
the outermost psymtab. However, filename searching doesn't (and
probably shouldn't) guarantee this. The fix is to walk upward to find
the outermost CU.
A new test case is included. It is mostly copied from other test
cases, which really sped up the effort.
This bug does not occur on trunk. There,
psym_map_symtabs_matching_filename is gone, replaced by
psymbol_functions::expand_symtabs_matching. When this find a match,
it calls psymtab_to_symtab, which does this same upward walk.
Tested on x86-64 Fedora 32.
I propose checking in this patch on the gdb-10 branch, and just the
new test case on trunk.
gdb/ChangeLog
2021-04-23 Tom Tromey <tromey@adacore.com>
gdb/testsuite/ChangeLog
2021-04-23 Tom Tromey <tromey@adacore.com>
PR gdb/27743:
* gdb.dwarf2/imported-unit-bp.exp: New file.
* gdb.dwarf2/imported-unit-bp-main.c: New file.
* gdb.dwarf2/imported-unit-bp-alt.c: New file.
Simon Marchi [Thu, 22 Apr 2021 19:05:18 +0000 (15:05 -0400)]
gdb: fix getting range of flexible array member in Python
As reported in bug 27757, we get an internal error when doing:
$ cat test.c
struct foo {
int len;
int items[];
};
struct foo *p;
int main() {
return 0;
}
$ gcc test.c -g -O0 -o test
$ ./gdb -q -nx --data-directory=data-directory ./test -ex 'python gdb.parse_and_eval("p").type.target()["items"].type.range()'
Reading symbols from ./test...
/home/simark/src/binutils-gdb/gdb/gdbtypes.h:435: internal-error: LONGEST dynamic_prop::const_val() const: Assertion `m_kind == PROP_CONST' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)
This is because the Python code (typy_range) blindly reads the high
bound of the type of `items` as a constant value. Since it is a
flexible array member, it has no high bound, the property is undefined.
Since commit 8c2e4e0689 ("gdb: add accessors to struct dynamic_prop"),
the getters check that you are not getting a property value of the wrong
kind, so this causes a failed assertion.
Fix it by checking if the property is indeed a constant value before
accessing it as such. Otherwise, use 0. This restores the previous GDB
behavior: because the structure was zero-initialized, this is what was
returned before. But now this behavior is explicit and not accidental.
Add a test, gdb.python/flexible-array-member.exp, that is derived from
gdb.base/flexible-array-member.exp. It tests the same things, but
through the Python API. It also specifically tests getting the range
from the various kinds of flexible array member types (AFAIK it wasn't
possible to do the equivalent through the CLI).
gdb/ChangeLog:
PR gdb/27757
* python/py-type.c (typy_range): Check that bounds are constant
before accessing them as such.
* guile/scm-type.c (gdbscm_type_range): Likewise.
gdb/testsuite/ChangeLog:
PR gdb/27757
* gdb.python/flexible-array-member.c: New test.
* gdb.python/flexible-array-member.exp: New test.
* gdb.guile/scm-type.exp (test_range): Add test for flexible
array member.
* gdb.guile/scm-type.c (struct flex_member): New.
(main): Use it.
This is a recurring problem that exposes a design issue in the DWARF
per-BFD sharing feature. Things work well when loading a binary with
the same method (with/without index, with/without readnow) twice in a
row. But they don't work so well when loading a binary with different
methods. See this previous fix, for example:
efb763a5ea35 ("gdb: check for partial symtab presence in dwarf2_initialize_objfile")
That one handled the case where the first load is normal (uses partial
symbols) and the second load uses an index.
The problem is that when loading an objfile with a method A, we create a
dwarf2_per_bfd and some dwarf2_per_cu_data and initialize them with the
data belonging to that method. When loading another obfile sharing the
same BFD but with a different method B, it's not clear how to re-use the
dwarf2_per_bfd/dwarf2_per_cu_data previously created, because they
contain the data specific to method A.
I think the most sensible fix would be to not share a dwarf2_per_bfd
between two objfiles loaded with different methods. That means that two
objfiles sharing the same BFD and loaded the same way would share a
dwarf2_per_bfd. Two objfiles sharing the same BFD but loaded with
different methods would use two different dwarf2_per_bfd structures.
However, this isn't a trivial change. So to fix the known issue quickly
(including in the gdb 10 branch), this patch just disables all
dwarf2_per_bfd sharing for objfiles using READNOW.
Generalize the gdb.base/index-cache-load-twice.exp test to test all
the possible combinations of loading a file with partial symtabs, index
and readnow. Move it to gdb.dwarf2, since it really exercises features
of the DWARF reader.
gdb/ChangeLog:
PR gdb/27541
* dwarf2/read.c (dwarf2_has_info): Don't share dwarf2_per_bfd
with objfiles using READNOW.
projects / thirdparty / binutils-gdb.git / log
