]>
git.ipfire.org Git - thirdparty/freeradius-server.git/log
Matthew Newton [Wed, 26 Oct 2022 20:46:07 +0000 (21:46 +0100)]
CI-DEB: use "sid" instead of "bookworm" for NR debian repos
Matthew Newton [Wed, 26 Oct 2022 17:21:45 +0000 (18:21 +0100)]
CI: Run on GitHub unless it's the main FreeRADIUS repo
Alan T. DeKok [Wed, 26 Oct 2022 16:28:26 +0000 (12:28 -0400)]
manually enable / disable new conditions for tests
Alan T. DeKok [Tue, 25 Oct 2022 20:57:00 +0000 (16:57 -0400)]
SQL uses SQL-Group, so use the old conditions here, too
Alan T. DeKok [Tue, 25 Oct 2022 20:33:34 +0000 (16:33 -0400)]
upcast to 64-bits for integers
Alan T. DeKok [Tue, 25 Oct 2022 20:26:21 +0000 (16:26 -0400)]
typo
Matthew Newton [Tue, 25 Oct 2022 22:45:19 +0000 (23:45 +0100)]
CI: run sanitizers in a separate non-Docker job on GitHub
Matthew Newton [Tue, 25 Oct 2022 23:01:38 +0000 (00:01 +0100)]
CI: fixes for when running directly on GitHub images
James Jones [Tue, 25 Oct 2022 22:22:44 +0000 (17:22 -0500)]
Fix dangling pointer issue (#4786)
clang may not realize that memcpy() doesn't keep the pointers handed
to it around, so there's no risk of the address of an auto being kept
past the caller's return, hence the "dangling pointer" warning.
Instead, head it off at the pass by immediately returning if len
is zero, so in needn't be set to the address of an auto; the only
explicit passing of NULL for in passes 0 for len. (Comments explain
how if len == 0, nothing changes in the context.) One could argue
for checking in rather than or in addition to checking len... OTOH,
shouldn't passing NULL and a non-zero len break loudly?
Matthew Newton [Tue, 25 Oct 2022 16:39:34 +0000 (17:39 +0100)]
CI: disable sanitizers until we can work out why it deadlocks in Docker
Matthew Newton [Tue, 25 Oct 2022 14:14:16 +0000 (15:14 +0100)]
CI: fix comment
Arran Cudbard-Bell [Tue, 25 Oct 2022 13:21:49 +0000 (09:21 -0400)]
Only run the main CI build on our infrastructure
Nick Porter [Tue, 25 Oct 2022 10:09:36 +0000 (11:09 +0100)]
Remove unused
Alan T. DeKok [Mon, 24 Oct 2022 21:31:12 +0000 (17:31 -0400)]
push xlat condition parsing until later
to avoid having unused xlats hanging around when we parse
policies.
Alan T. DeKok [Mon, 24 Oct 2022 21:30:19 +0000 (17:30 -0400)]
register function arguments if available
so that the parser can do the correct thing
Alan T. DeKok [Mon, 24 Oct 2022 19:43:35 +0000 (15:43 -0400)]
add "don't use new conditions" for LDAP tests
so that when we "flip the switch" to always enable them, nothing
should break
Nick Porter [Mon, 24 Oct 2022 08:58:56 +0000 (09:58 +0100)]
Amend Active Directory persistent search tests for parallel runs
Nick Porter [Mon, 24 Oct 2022 08:08:52 +0000 (09:08 +0100)]
Amend persistent search tests to use different output files for parallel runs
Nick Porter [Mon, 24 Oct 2022 08:07:34 +0000 (09:07 +0100)]
Amend persistent search tests to use new entries
Nick Porter [Mon, 24 Oct 2022 08:05:13 +0000 (09:05 +0100)]
Add extra entries to 389ds test directory
Matthew Newton [Sun, 23 Oct 2022 17:27:06 +0000 (18:27 +0100)]
CI: clean up docker cruft after job runs
Arran Cudbard-Bell [Sat, 22 Oct 2022 20:21:51 +0000 (16:21 -0400)]
Make the same fix for mkdirat
Arran Cudbard-Bell [Sat, 22 Oct 2022 20:12:47 +0000 (16:12 -0400)]
If the directory already exists then we need to provide a handle to it
Arran Cudbard-Bell [Sat, 22 Oct 2022 19:03:12 +0000 (15:03 -0400)]
Return the actual number of bytes processed, and the actual negative failure offset in fd_mkdir
Arran Cudbard-Bell [Sat, 22 Oct 2022 18:22:25 +0000 (14:22 -0400)]
Returning path - p where p is NULL results in a positive value
Arran Cudbard-Bell [Sat, 22 Oct 2022 17:50:10 +0000 (13:50 -0400)]
Unclear at this point why mkdirat is returning EBADF
Nick Porter [Sat, 22 Oct 2022 15:42:36 +0000 (16:42 +0100)]
Allow FR to start its LDAP persistent search before starting tests
Nick Porter [Sat, 22 Oct 2022 15:42:54 +0000 (16:42 +0100)]
Delete correct output file
Nick Porter [Sat, 22 Oct 2022 17:33:49 +0000 (18:33 +0100)]
Revert "Re-work ldap setup script"
ss is not available on self-hosted runners, and further
observation shows this is not the cause of the random
failures
Arran Cudbard-Bell [Sat, 22 Oct 2022 17:23:50 +0000 (13:23 -0400)]
Need to check for both...
Arran Cudbard-Bell [Sat, 22 Oct 2022 17:00:29 +0000 (13:00 -0400)]
*sigh* the real issue is path is not const
Arran Cudbard-Bell [Sat, 22 Oct 2022 16:51:11 +0000 (12:51 -0400)]
const
Dominik George [Sat, 22 Oct 2022 16:49:05 +0000 (18:49 +0200)]
Fix comment on status line length in rlm_rest (#4783)
Arran Cudbard-Bell [Sat, 22 Oct 2022 16:46:25 +0000 (12:46 -0400)]
Don't return garbage when failing to create the top level directory
Arran Cudbard-Bell [Sat, 22 Oct 2022 08:55:08 +0000 (04:55 -0400)]
Don't lose the context for mkdir errors
Arran Cudbard-Bell [Sat, 22 Oct 2022 08:52:07 +0000 (04:52 -0400)]
Remove one fork() per utility invocation
Arran Cudbard-Bell [Sat, 22 Oct 2022 08:21:02 +0000 (04:21 -0400)]
It's fine if the dir we're making already exists
Arran Cudbard-Bell [Sat, 22 Oct 2022 08:16:29 +0000 (04:16 -0400)]
syserror isn't useful here, whereas our thread local buffer will be
Arran Cudbard-Bell [Sat, 22 Oct 2022 08:15:33 +0000 (04:15 -0400)]
Catch commands returning invalid codes
Arran Cudbard-Bell [Sat, 22 Oct 2022 07:34:52 +0000 (03:34 -0400)]
Pass mode when creating output files
Arran Cudbard-Bell [Sat, 22 Oct 2022 07:25:16 +0000 (03:25 -0400)]
Need sys/file.h
Arran Cudbard-Bell [Sat, 22 Oct 2022 07:19:56 +0000 (03:19 -0400)]
Prevent several race conditions in unit_test_attribute
Make writing as robust as possible
Nick Porter [Fri, 21 Oct 2022 20:07:42 +0000 (21:07 +0100)]
Move to node16 version of github actions
Nick Porter [Fri, 21 Oct 2022 19:23:15 +0000 (20:23 +0100)]
Remove deprecated set-output
Nick Porter [Fri, 21 Oct 2022 18:53:45 +0000 (19:53 +0100)]
Re-work ldap setup script
To check server is listening before attempting to add data
Arran Cudbard-Bell [Fri, 21 Oct 2022 19:02:25 +0000 (15:02 -0400)]
...and no one actually put the sneaky alias in FreeBSD
Arran Cudbard-Bell [Fri, 21 Oct 2022 18:57:14 +0000 (14:57 -0400)]
Except when a test sneakily creates a wrapper script
but then not so sneakily fails to catch all the previous sysctl instances
Alan T. DeKok [Tue, 18 Oct 2022 14:10:48 +0000 (10:10 -0400)]
signal the request to cancel, too
Arran Cudbard-Bell [Fri, 21 Oct 2022 18:53:06 +0000 (14:53 -0400)]
nproc doesn't exist on freebsd
Terry Burton [Fri, 21 Oct 2022 18:41:30 +0000 (19:41 +0100)]
test: Quick wins by performing some tests in parallel (#4782)
Nick Porter [Fri, 21 Oct 2022 16:40:42 +0000 (17:40 +0100)]
Update tests to use different output filenames for parallel running
Nick Porter [Fri, 21 Oct 2022 16:39:45 +0000 (17:39 +0100)]
Amend tests to use new entries
Nick Porter [Fri, 21 Oct 2022 16:37:22 +0000 (17:37 +0100)]
Add extra entries to LDAP directory for parallel tests
James Jones [Fri, 21 Oct 2022 17:59:13 +0000 (12:59 -0500)]
Ignore returns from certain sbuff calls (CIDs #
1504025 , #
1504070 ) (#4750)
These are cases in which sbuffs are either created from C arrays
or are allocated from the heap but given initial and maximum sizes.
The author(s) of the code presumably chose sizes intended to suffice,
but despite that, coverity will notice that the sbuff functions
used return a value, and hence will report check_return barring
either explicit cast to void or annotation. We therefore cast to void.
James Jones [Fri, 21 Oct 2022 17:58:55 +0000 (12:58 -0500)]
Check return of fr_pair_list_copy_by_da() (CID #
1469148 ) (#4752)
James Jones [Fri, 21 Oct 2022 17:51:58 +0000 (12:51 -0500)]
Explicitly ignore the curl_mime_type() return (CID #
1504071 ) (#4764)
* Check curl calls that allocate memory (CID #
1504071 )
This includes the curl_mime_type() call that gives rise to the
coverity defect, but also checks the mime and mime part allocations
that must work for body_init() to work.
* Formatting
Co-authored-by: Arran Cudbard-Bell <a.cudbardb@freeradius.org>
James Jones [Fri, 21 Oct 2022 17:50:25 +0000 (12:50 -0500)]
Check fr_event_timer_in() return value (CID #
1503926 ) (#4774)
* Check fr_event_timer_in() return value (CID #
1503926 )
* Assert and error if we can't insert timer events
Co-authored-by: Arran Cudbard-Bell <a.cudbardb@freeradius.org>
Terry Burton [Fri, 21 Oct 2022 17:46:39 +0000 (18:46 +0100)]
tests: Ignore spurious output from jlibtool with VERBOSE=1 (#4781)
Arran Cudbard-Bell [Thu, 20 Oct 2022 02:40:34 +0000 (22:40 -0400)]
Add ability to set static headers to rlm_rest
Max Khon [Wed, 19 Oct 2022 13:36:49 +0000 (16:36 +0300)]
fr_value_box_from_network: Fix NULL pointer dereference when truncating (#4779)
James Jones [Tue, 18 Oct 2022 17:15:55 +0000 (12:15 -0500)]
Annotate fs_check_call (CID #
1271307 ) (#4778)
The unlink() call (the use of the toctou) does check its return
code. Also, it's not liwted among the UseSet functions in "TOCTTOU
Vulnerabilities in Unix-Style File Systems: An Anatomical Study",
https://www.usenix.org/legacy/events/fast05/tech/full_papers/wei/wei.pdf
James Jones [Tue, 18 Oct 2022 17:15:35 +0000 (12:15 -0500)]
Ignore fr_sbuff_in_strcpy() return in fr_pair_list_log_sbuff() (CID #
1504278 ) (#4775)
sbuff is emptied each time through the for loop, and down this
path it will only contain the output of fr_dict_attr_oid_print(),
possibly preceded by "raw.". fr_pair_list_log_sbuff() is only
called from one place, which passes in an sbuff using an array
of 1024 characters. This should suffice to make appending " = "
always work.
James Jones [Tue, 18 Oct 2022 17:13:07 +0000 (12:13 -0500)]
Ignore fr_event_timer_in() return here, too (CID #154036) (#4773)
Like
1503936 , this is in a callback and will reuse the idle
timeout event.
James Jones [Tue, 18 Oct 2022 17:12:55 +0000 (12:12 -0500)]
Deal with time of check/time of use complaints (CIDs below) (#4768)
1400053 : the call coverity complains about is just there to
determine why the dlopen() failed.
1445217 : the fopen() return is checked, so any changes between
the stat() call and it should be detected.
1503910 : here, the unlink() return is checked
Alan T. DeKok [Tue, 18 Oct 2022 13:57:49 +0000 (09:57 -0400)]
only NDEBUG
Arran Cudbard-Bell [Tue, 18 Oct 2022 13:31:34 +0000 (09:31 -0400)]
Add username to the REDIS auth command
Alan T. DeKok [Tue, 18 Oct 2022 12:53:35 +0000 (08:53 -0400)]
typo
Alan T. DeKok [Tue, 18 Oct 2022 12:44:01 +0000 (08:44 -0400)]
cleanups and fixes for #4777
Move common macros and includes to a common header file.
Create a macro which wraps OpenBSD's closefrom() in a macro,
so that the source code isn't littered with ifdef's
Alan T. DeKok [Tue, 18 Oct 2022 12:31:17 +0000 (08:31 -0400)]
don't run this in parallel
Alan T. DeKok [Mon, 17 Oct 2022 20:09:49 +0000 (16:09 -0400)]
update request_clone() so that it works
Alan T. DeKok [Mon, 17 Oct 2022 19:43:22 +0000 (15:43 -0400)]
do perf updates before freeing the bits from the frame
Alan T. DeKok [Mon, 17 Oct 2022 19:12:36 +0000 (15:12 -0400)]
add "interpret" section with instruction limiting pieces
Alan T. DeKok [Mon, 17 Oct 2022 19:03:20 +0000 (15:03 -0400)]
add maximum instruction count for failure testing
Matthew Newton [Sun, 16 Oct 2022 18:35:54 +0000 (19:35 +0100)]
CI: bump timeout
see if that's causing random failures
Christian Hesse [Sun, 16 Oct 2022 11:39:24 +0000 (13:39 +0200)]
Fallback to `uname -n` on missing `hostname` in radtest (#4771)
This should work without inetutils being installed, so add a fallback.
Signed-off-by: Christian Hesse <mail@eworm.de>
Signed-off-by: Christian Hesse <mail@eworm.de>
Matthew Newton [Fri, 14 Oct 2022 16:48:25 +0000 (17:48 +0100)]
CI: remove debugging
Matthew Newton [Fri, 14 Oct 2022 14:47:04 +0000 (15:47 +0100)]
CI: install git-lfs early, too
James Jones [Fri, 14 Oct 2022 14:14:44 +0000 (09:14 -0500)]
Explicitly ignore fr_event_timer_in() return value (CID #
1503936 ) (#4769)
fr_event_timer_in() (actually _fr_event_timer_at()) reuses events
where possible, and being in the idle timeout means that there is
such an event available for reuse.
Matthew Newton [Fri, 14 Oct 2022 12:32:05 +0000 (13:32 +0100)]
CI: Tidy home directory in Docker
Matthew Newton [Thu, 13 Oct 2022 16:18:40 +0000 (17:18 +0100)]
CI: ruby/setup-ruby broken in GitHub container right now
Matthew Newton [Thu, 13 Oct 2022 14:45:22 +0000 (15:45 +0100)]
CI: for the time being run on both GitHub and self-hosted
Matthew Newton [Wed, 12 Oct 2022 23:03:50 +0000 (00:03 +0100)]
CI: fixups to use 389ds in Docker container
Matthew Newton [Mon, 10 Oct 2022 13:34:35 +0000 (14:34 +0100)]
CI: use vars to set llvm/gcc versions
Matthew Newton [Thu, 13 Oct 2022 14:02:19 +0000 (15:02 +0100)]
CI: move fuzzer tests to separate composite action
Matthew Newton [Thu, 13 Oct 2022 11:33:47 +0000 (12:33 +0100)]
CI: move main tests to separate composite action
Matthew Newton [Wed, 12 Oct 2022 22:47:02 +0000 (23:47 +0100)]
CI: move build process to separate composite action
Matthew Newton [Wed, 12 Oct 2022 22:54:22 +0000 (23:54 +0100)]
CI: prepare Docker image more like GitHub runner
Matthew Newton [Wed, 12 Oct 2022 21:52:15 +0000 (22:52 +0100)]
CI: run in Docker and add test service containers
- mariadb
- postgres
- redis
- 389ds
Matthew Newton [Wed, 12 Oct 2022 19:58:37 +0000 (20:58 +0100)]
CI: add matrix.os to set where jobs run
Matthew Newton [Wed, 12 Oct 2022 19:57:40 +0000 (20:57 +0100)]
CI: move macOS to a separate job
Matthew Newton [Thu, 13 Oct 2022 13:48:03 +0000 (14:48 +0100)]
CI: actions/setup-ruby is deprecated
Matthew Newton [Mon, 10 Oct 2022 15:18:00 +0000 (16:18 +0100)]
CI: explicitly enable developer mode
Matthew Newton [Thu, 13 Oct 2022 09:58:42 +0000 (10:58 +0100)]
CI: fix YAML whitespace
Jorge Pereira [Thu, 13 Oct 2022 02:16:34 +0000 (23:16 -0300)]
rlm_sql_oracle: Fix typo (#4765)
Jorge Pereira [Wed, 12 Oct 2022 12:50:00 +0000 (09:50 -0300)]
Suppress MacOS ld warning from native compilation (#4766)
Add the option '-Wl,-w' as suggested in https://issues.guix.gnu.org/issue/57849
Nick Porter [Wed, 12 Oct 2022 10:37:17 +0000 (11:37 +0100)]
Improve queries for processing radacct into periodic usage data
Removing `OR` from the query allows better index usage
Nick Porter [Wed, 12 Oct 2022 10:32:55 +0000 (11:32 +0100)]
Typo
Jorge Pereira [Tue, 11 Oct 2022 20:54:30 +0000 (17:54 -0300)]
Fix typo (#4763)
Jorge Pereira [Thu, 6 Oct 2022 17:45:37 +0000 (14:45 -0300)]
eapol_test: Add paths for Brew on OSX/M1 (#4757)
Needed for the Brew running on OSX/M1 due to the new base path is in /opt/homebrew
as described on https://earthly.dev/blog/homebrew-on-m1/
Arran Cudbard-Bell [Thu, 6 Oct 2022 04:02:52 +0000 (00:02 -0400)]
Make plain 'autoconf' 'autoheader' 'autoupdate' invocations work again
Reading the autoconf docs the preferred way to do includes is to set `AC_CONFIG_MACRO_DIR([m4])` then run aclocal to figure out which .m4 files need including.
Check the size of time_t so we can determine if we're going to overflow it...