Hi Mark,
On Thu, Oct 22, 2026 at 11:20:44PM -0700, Mark Harris wrote:
> I encountered 50 regtest failures running Valgrind 3.28.0.RC1
> 1ab15ca91e86 on Fedora Linux 44 (COSMIC) aarch64.
That is a bit more than we like, but it seems most can be explained.
> == 795 tests, 48 stderr failures, 1 stdout failure, 0 stderrB
> failures, 1 stdoutB failure, 0 post failures ==
>
> The failures fall into a few categories.
>
> 21 tests failed due to extra inherited fds from the desktop
> environment; these pass if rerun outside of the desktop environment:
> none/tests/fdleak_cmsg (stderr)
> none/tests/fdleak_cmsg_supp (stderr)
> none/tests/fdleak_cmsg_xml (stderr)
> none/tests/fdleak_creat (stderr)
> none/tests/fdleak_creat_sup (stderr)
> none/tests/fdleak_creat_xml (stderr)
> none/tests/fdleak_dup (stderr)
> none/tests/fdleak_dup2 (stderr)
> none/tests/fdleak_dup2_xml (stderr)
> none/tests/fdleak_dup_xml (stderr)
> none/tests/fdleak_fcntl (stderr)
> none/tests/fdleak_fcntl_xml (stderr)
> none/tests/fdleak_ipv4 (stderr)
> none/tests/fdleak_open (stderr)
> none/tests/fdleak_pipe (stderr)
> none/tests/fdleak_socketpair (stderr)
> none/tests/linux/getdents_filter (stdout)
> none/tests/track_high (stderr)
> none/tests/track_yes (stderr)
> none/tests/rlimit64_nofile (stderr)
> none/tests/rlimit_nofile (stderr)
That probably means there is some bug (file descriptor leak) in your
desktop environment, it probably launches programs with some still
open file descriptor which then gets inherited.
> 10 failed because "Max number of guard pages is too low":
> memcheck/tests/err_disable4 (stderr)
> memcheck/tests/thread_alloca (stderr)
> helgrind/tests/annotate_smart_pointer (stderr)
> helgrind/tests/pth_barrier3 (stderr)
> drd/tests/annotate_smart_pointer (stderr)
> drd/tests/concurrent_close (stderr)
> drd/tests/pth_barrier3 (stderr)
> drd/tests/pth_barrier_thr_cr (stderr)
> drd/tests/pth_create_chain (stderr)
> drd/tests/swapcontext (stderr)
I don't understand these. Would you be able to recompile with this
patch:
diff --git a/coregrind/m_aspacemgr/aspacemgr-linux.c
b/coregrind/m_aspacemgr/aspacemgr-linux.c
index 6ede41a0151f..a9ee0e3e670d 100644
--- a/coregrind/m_aspacemgr/aspacemgr-linux.c
+++ b/coregrind/m_aspacemgr/aspacemgr-linux.c
@@ -1107,7 +1107,7 @@ static void guard_page_install ( Addr addr ) {
Addr addr_aligned = addr & ~(VKI_PAGE_SIZE - 1);
if (nguardpages_used >= VG_N_GUARDS) {
VG_(printf)("Use --max-guard-pages=INT to specify a larger number of\n"
- "guard pages and rerun valgrind\n");
+ "guard pages (current %u) and rerun valgrind\n",
VG_N_GUARDS);
VG_(core_panic)("Max number of guard pages is too low");
}
// bisect
We might be not be initializing VG_N_GUARDS correctly. Or something
in your environment installs lots (500+?) guardpages for some reason.
> 8 are missing a line in the backtrace:
> memcheck/tests/fwrite (stderr)
> memcheck/tests/linux/sys-copy_file_range (stderr)
> memcheck/tests/linux/sys-preadv_pwritev (stderr)
> memcheck/tests/sendmsg (stderr)
> memcheck/tests/writev1 (stderr)
> none/tests/track_bad (stderr)
> none/tests/track_new (stderr)
> none/tests/use_after_close (stderr)
>
> 7 have different varinfo descriptions:
> memcheck/tests/dw4 (stderr)
> memcheck/tests/varinfo2 (stderr)
> memcheck/tests/varinfo4 (stderr)
> memcheck/tests/varinfo5 (stderr)
> memcheck/tests/varinfo6 (stderr)
> memcheck/tests/varinforestrict (stderr)
> helgrind/tests/hg05_race2 (stderr)
These are also seen on other arm64 setups. We seem to not
correctly/exactly unwind.
> 4 have other failures:
> gdbserver_tests/hgtls (stdoutB)
> helgrind/tests/tc20_verifywrap (stderr)
> drd/tests/pth_detached3 (stderr)
> drd/tests/tls_threads (stderr)
These also (sometimes) fail on other arches.
> Full regtest diffs: https://paste.centos.org/view/95f06742
Thanks,
Mark
_______________________________________________
Valgrind-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/valgrind-users
