Bug#936055: t/unthreaded.t test failing
Control: reassign -1 gdb 9.1-1
Control: forwarded -1
https://sourceware.org/ml/gdb-patches/2020-03/msg00010.html
Control: retitle -1 [regression] GDB can't printf convenience vars that hold
pointers to strings
On Wednesday, February 19 2020, gregor herrmann wrote:
> On Wed, 19 Feb 2020 14:16:52 +0100, Héctor Orón Martínez wrote:
>
>> This issue is now affecting GDB 9.1 autopkgtests and preventing it from
>> migrating to testing.
>>
>> https://ci.debian.net/data/autopkgtest/testing/amd64/liba/libapp-stacktrace-perl/4328197/log.gz
>>
>> https://ci.debian.net/data/autopkgtest/testing/arm64/liba/libapp-stacktrace-perl/4314744/log.gz
>
> Oh, one of the bugs which "hide":
> https://ci.debian.net/packages/liba/libapp-stacktrace-perl/testing/amd64/
> is all green, and after clicking on britney it turns red:
> https://ci.debian.net/user/britney/jobs?package=libapp-stacktrace-perl[]=testing[]=amd64
Hi there,
After spending a non-trivial amount of time investigating this over the
weekend, I was able to find the underlying case of the bug (and come up
with a fix). Please, keep reading.
> Anyway, in the meantime it also FTBFS in unstable:
>
> # [Thread debugging using libthread_db enabled]
> # Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
> # 0x7f7b77cc1047 in select () from /lib/x86_64-linux-gnu/libc.so.6
> # à|ÚU:101
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # [Inferior 1 (process 25884) detached]
>
> # Failed test at t/unthreaded.t line 55.
> # '[Thread debugging using libthread_db enabled]
> # Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
> # 0x7f7b77cc1047 in select () from /lib/x86_64-linux-gnu/libc.so.6
> # à|ÚU:101
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # à|ÚU:106
> # [Inferior 1 (process 25884) detached]
> # '
> # doesn't match '(?^mx:
> # (?:
> # ^t/unthreaded\.t:\d+\n
> # ){10}
> # )'
> # Looks like you failed 1 test of 5.
> t/unthreaded.t
> 1..5
> not ok 1
> ok 2 - Reaped pstack
> ok 3 - exit(0)
> ok 4 - No signals
> ok 5 - No core dump
> # Alive top 1582127930: 1 middle: 1
> # Alive top 1582127931: 1 middle: 0
> Dubious, test returned 1 (wstat 256, 0x100)
> Failed 1/5 subtests
>
> Test Summary Report
> ---
> t/unthreaded.t (Wstat: 256 Tests: 5 Failed: 1)
> Failed test: 1
> Non-zero exit status: 1
> Files=4, Tests=6, 2 wallclock secs ( 0.03 usr 0.01 sys + 0.96 cusr 0.09
> csys = 1.09 CPU)
> Result: FAIL
>
> (I _think_ that's yet another different error.)
This seems to be the same failure reported by doko.
> In testing with gdb 8.3.1 the package still builds:
>
> # [Thread debugging using libthread_db enabled]
> # Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
> # 0x7f1b9eb77047 in select () from /lib/x86_64-linux-gnu/libc.so.6
> # t/unthreaded.t:101
> # t/unthreaded.t:106
> # t/unthreaded.t:106
> # t/unthreaded.t:106
> # t/unthreaded.t:106
> # t/unthreaded.t:106
> # t/unthreaded.t:106
> # t/unthreaded.t:106
> # t/unthreaded.t:106
> # t/unthreaded.t:106
> # t/unthreaded.t:106
> # [Inferior 1 (process 3207) detached]
> t/unthreaded.t
> 1..5
> ok 1
> ok 2 - Reaped pstack
> ok 3 - exit(0)
> ok 4 - No signals
> ok 5 - No core dump
> # Alive top 1582128303: 1 middle: 1
> # Alive top 1582128304: 1 middle: 0
> ok
Right. This is actually a regression that happened on GDB (therefore
I'm reassigning the bug).
So, it took me quite a bit of time to debug this because
libapp-stacktrace-perl doesn't help when it comes to seeing what GDB is
doing inside it... But I was able to determine that the problem started
happening after the following upstream commit:
commit 1f6f6e21fa86dc3411a6498608f32e9eb24b7851
Author: Philippe Waroquiers
Date: Mon Jun 10 21:41:51 2019 +0200
Ensure GDB printf command can print convenience var strings without a
target.
As the commit title says, the goal here was to make GDB's "printf"
command be able to print convenience variables that are strings without
the need for a target (a program/corefile being debugged, for example).
However, the commit changed the way GDB treats convenience variables
that *point* to addresses referring to e.g. a "char *" inside the
debuggee's memory space. GDB will treat this as a string (because it's
pointing to a "char *"), and will try to obtain its value via the wrong
mechanism.
I won't dive too deep here; suffice it to say that if we treat
convenience variables that hold pointers in a different manner, then
things get back to normal.
I submitted a tentative patch to fix this upstream, and will backport it
downstream once my approach is deemed correct.
Thanks,
--
Sergio
GPG key ID: 237A 54B1 0287 28BF 00EF
Bug#936055: t/unthreaded.t test failing
Control: severity -1 serious
On Wed, 19 Feb 2020 14:16:52 +0100, Héctor Orón Martínez wrote:
> This issue is now affecting GDB 9.1 autopkgtests and preventing it from
> migrating to testing.
>
> https://ci.debian.net/data/autopkgtest/testing/amd64/liba/libapp-stacktrace-perl/4328197/log.gz
>
> https://ci.debian.net/data/autopkgtest/testing/arm64/liba/libapp-stacktrace-perl/4314744/log.gz
Oh, one of the bugs which "hide":
https://ci.debian.net/packages/liba/libapp-stacktrace-perl/testing/amd64/
is all green, and after clicking on britney it turns red:
https://ci.debian.net/user/britney/jobs?package=libapp-stacktrace-perl[]=testing[]=amd64
Anyway, in the meantime it also FTBFS in unstable:
# [Thread debugging using libthread_db enabled]
# Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
# 0x7f7b77cc1047 in select () from /lib/x86_64-linux-gnu/libc.so.6
# à|ÚU:101
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# [Inferior 1 (process 25884) detached]
# Failed test at t/unthreaded.t line 55.
# '[Thread debugging using libthread_db enabled]
# Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
# 0x7f7b77cc1047 in select () from /lib/x86_64-linux-gnu/libc.so.6
# à|ÚU:101
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# à|ÚU:106
# [Inferior 1 (process 25884) detached]
# '
# doesn't match '(?^mx:
# (?:
# ^t/unthreaded\.t:\d+\n
# ){10}
# )'
# Looks like you failed 1 test of 5.
t/unthreaded.t
1..5
not ok 1
ok 2 - Reaped pstack
ok 3 - exit(0)
ok 4 - No signals
ok 5 - No core dump
# Alive top 1582127930: 1 middle: 1
# Alive top 1582127931: 1 middle: 0
Dubious, test returned 1 (wstat 256, 0x100)
Failed 1/5 subtests
Test Summary Report
---
t/unthreaded.t (Wstat: 256 Tests: 5 Failed: 1)
Failed test: 1
Non-zero exit status: 1
Files=4, Tests=6, 2 wallclock secs ( 0.03 usr 0.01 sys + 0.96 cusr 0.09
csys = 1.09 CPU)
Result: FAIL
(I _think_ that's yet another different error.)
In testing with gdb 8.3.1 the package still builds:
# [Thread debugging using libthread_db enabled]
# Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
# 0x7f1b9eb77047 in select () from /lib/x86_64-linux-gnu/libc.so.6
# t/unthreaded.t:101
# t/unthreaded.t:106
# t/unthreaded.t:106
# t/unthreaded.t:106
# t/unthreaded.t:106
# t/unthreaded.t:106
# t/unthreaded.t:106
# t/unthreaded.t:106
# t/unthreaded.t:106
# t/unthreaded.t:106
# t/unthreaded.t:106
# [Inferior 1 (process 3207) detached]
t/unthreaded.t
1..5
ok 1
ok 2 - Reaped pstack
ok 3 - exit(0)
ok 4 - No signals
ok 5 - No core dump
# Alive top 1582128303: 1 middle: 1
# Alive top 1582128304: 1 middle: 0
ok
Cheers,
gregor
--
.''`. https://info.comodo.priv.at -- Debian Developer https://www.debian.org
: :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D 85FA BB3A 6801 8649 AA06
`. `' Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation Europe
`- NP: Pink Floyd: Another Brick In The Wall (part 1)
signature.asc
Description: Digital Signature
Bug#936055: t/unthreaded.t test failing
Source: libapp-stacktrace-perl Version: 0.09-3 Followup-For: Bug #936055 Dear Maintainer, This issue is now affecting GDB 9.1 autopkgtests and preventing it from migrating to testing. https://ci.debian.net/data/autopkgtest/testing/amd64/liba/libapp-stacktrace-perl/4328197/log.gz https://ci.debian.net/data/autopkgtest/testing/arm64/liba/libapp-stacktrace-perl/4314744/log.gz Regards -- System Information: Debian Release: bullseye/sid APT prefers unstable APT policy: (500, 'unstable'), (1, 'experimental') Architecture: amd64 (x86_64) Foreign Architectures: armhf Kernel: Linux 5.3.0-1-amd64 (SMP w/8 CPU cores) Kernel taint flags: TAINT_PROPRIETARY_MODULE, TAINT_OOT_MODULE, TAINT_UNSIGNED_MODULE Locale: LANG=ca_AD.UTF-8, LC_CTYPE=ca_AD.UTF-8 (charmap=UTF-8), LANGUAGE=ca_AD:ca (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Init: systemd (via /run/systemd/system) LSM: AppArmor: enabled
Bug#936055: t/unthreaded.t test failing
On Sat, 31 Aug 2019 14:57:15 +0200, gregor herrmann wrote:
> Some random notes:
What I also noted, probably unrelated but still:
During the build:
dpkg-gencontrol: warning: Depends field of package perl-stacktrace:
substitution variable ${shlibs:Depends} used, but is not defined
dpkg-gencontrol: warning: Depends field of package perl-stacktrace:
substitution variable ${shlibs:Depends} used, but is not defined
Lintian:
W: perl-stacktrace: shared-lib-without-dependency-information
usr/lib/x86_64-linux-gnu/perl5/5.28/auto/App/Stacktrace/Stacktrace.so
(ldd indeed returns "static")
Cheers,
gregor
--
.''`. https://info.comodo.priv.at -- Debian Developer https://www.debian.org
: :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D 85FA BB3A 6801 8649 AA06
`. `' Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation Europe
`- NP: Lightnin' Hopkins: Cryin' Shame
signature.asc
Description: Digital Signature
Bug#936055: t/unthreaded.t test failing
On Thu, 29 Aug 2019 15:45:03 +0200, Matthias Klose wrote: > Package: src:libapp-stacktrace-perl > Version: 0.09-3 > Severity: important > Tags: upstream > Forwarded: https://rt.cpan.org/Public/Bug/Display.html?id=121262 > > I see you already forwarded this upstream. Filing the issue, because I also > see that in Ubuntu at > https://objectstorage.prodstack4-5.canonical.com/v1/AUTH_77e2ada1e7a84929a74ba3b87153c0ac/autopkgtest-eoan/eoan/amd64/liba/libapp-stacktrace-perl/20190828_222752_2da85@/log.gz Thanks for the bug report. Some random notes: - I can't get t/unthreaded.t to fail when running it in a loop (in a cowbuilder chroot) for hours. - The failure in the CPAN ticket looks different than the one from the Ubuntu autopkgtest. - The package has a patch [0] which mentions a bug [1] which also has a failure in t/unthreaded.t (which looks more like the CPAN failure than the Ubuntu failure). > Is there a way to exit with a status for unreliable tests, that gets ignored > by the autopkg tests? None that I'm aware of, at least with the autopkgtest-pkg-perl framework. And I suspect that the problem should show up not only for autopkgtests but also during build, at least potentially. But I note that it builds/tests fine about everywhere we have in Debian (buildds, reproducible builds, and also debci for testing and unstable just a few days ago [2]). Cheers, gregor, confused [0] https://sources.debian.org/src/libapp-stacktrace-perl/0.09-3/debian/patches/0001-Pointers-should-be-long-not-int.patch/ [1] https://bugs.debian.org/775744 [2] https://ci.debian.net/packages/liba/libapp-stacktrace-perl/testing/amd64/ https://ci.debian.net/packages/liba/libapp-stacktrace-perl/unstable/amd64/ -- .''`. https://info.comodo.priv.at -- Debian Developer https://www.debian.org : :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D 85FA BB3A 6801 8649 AA06 `. `' Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation Europe `- NP: Lightnin' Hopkins: Cryin' Shame signature.asc Description: Digital Signature
Bug#936055: t/unthreaded.t test failing
Package: src:libapp-stacktrace-perl Version: 0.09-3 Severity: important Tags: upstream Forwarded: https://rt.cpan.org/Public/Bug/Display.html?id=121262 I see you already forwarded this upstream. Filing the issue, because I also see that in Ubuntu at https://objectstorage.prodstack4-5.canonical.com/v1/AUTH_77e2ada1e7a84929a74ba3b87153c0ac/autopkgtest-eoan/eoan/amd64/liba/libapp-stacktrace-perl/20190828_222752_2da85@/log.gz Is there a way to exit with a status for unreliable tests, that gets ignored by the autopkg tests?
