Package: bash
Version: 5.1~rc2-1
Severity: normal
X-Debbugs-Cc: [email protected]
Dear Maintainer,
I'm running a system that tracks Debian testing somewhat closely (as you can
see it's a bit behind ;)). When running the test suite for Mercurial, at least
three of us have seen lines like these in the output, after a `wait` call:
```
TRACE: pid 3755293: bgp_delete: deleting 3755509
```
I believe I've tracked this down to a line in the `configure` script, which
has this block:
```
case "$RELSTATUS" in
alp*|bet*|dev*|rc*|releng*|maint*) DEBUG='-DDEBUG'
MALLOC_DEBUG='-DMALLOC_DEBUG' ;;
*) DEBUG= MALLOC_DEBUG= ;;
esac
```
Since Debian testing has a 'release candidate' version at the moment, it was
built with -DDEBUG and -DMALLOC_DEBUG. The first define causes it to produce
the TRACE messages, the second one presumably causes it to use a malloc that
might operate slower, leading to visible performance regressions.
I've not been able to reproduce by hand on the commandline, so it's entirely
possible that there's two bugs here, one in the Mercurial test suite (that
causes the TRACEd condition), and one in the packaging of bash (making it emit
that TRACE message). I believe they should be looked at separately, and that
the bash packaging issue is actually the more severe one.
I was unsure what severity level to put this at, I chose 'normal', due to a
non-rc version being available on sid already. I have three primary objectives
for filing:
1. raise the issue so that rc versions are not promoted beyond testing
2. question whether it makes sense to have rc versions on testing at all
(maybe they should be limited to unstable or experimental)
3. question whether the rc versions *should* be built differently from the
actual release versions, at least when built by distros. Caveat: I believe
that the system I'm running actually does its own builds of packages
before they show up on my machine; it's technically possible that this is
not actually an issue for Debian-provided rc binaries? I don't have
another machine to use to test that hypothesis.
I'm specifically *not* attempting to have the output removed via a code change
to bash (either to remove the output, or to find the reason for it and [if
it's a bug] fix it). This bug report is about the build/package/release
process, the exact message that's being printed is mostly irrelevant. I
primarily included it for searchability and reproducibility.
To reproduce:
```
$ hg clone https://mercurial-scm.org/repo/hg
$ cd hg/tests
$ ./run-tests.py test-pending.t
<several seconds delay while it builds hg>
running 1 tests using 1 parallel processes
--- /home/spectral/tmp/bugreport/hg/tests/test-debugcommands.t
+++ /home/spectral/tmp/bugreport/hg/tests/test-debugcommands.t.err
@@ -447,6 +447,7 @@
[1]
$ touch .hg/unlock
$ wait
+ TRACE: pid 3755293: bgp_delete: deleting 3755509
$ [ -f .hg/store/lock ] || echo "There is no lock"
There is no lock
... several other instances ...
```
Thank you for your time and consideration.
-- System Information:
Debian Release: rodete
Architecture: amd64 (x86_64)
Foreign Architectures: i386
Kernel: Linux 5.7.17-1rodete4-amd64 (SMP w/36 CPU threads)
Kernel taint flags: TAINT_PROPRIETARY_MODULE, TAINT_OOT_MODULE,
TAINT_UNSIGNED_MODULE
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8), LANGUAGE not set
Shell: /bin/sh linked to /bin/bash
Init: systemd (via /run/systemd/system)
LSM: AppArmor: enabled
Versions of packages bash depends on:
ii base-files 11+gl0
ii debianutils 4.11.2
ii libc6 2.31-3
ii libtinfo6 6.2+20200918-1
Versions of packages bash recommends:
ii bash-completion 1:2.11-2
Versions of packages bash suggests:
pn bash-doc <none>
-- Configuration Files:
/etc/bash.bashrc changed [not included]
/etc/skel/.bashrc changed [not included]
-- no debconf information
