Bug#1052942: insserv: FTBFS: insserv: Could not read script nolsbheader: No such file or directory

[Mark Hindley]

Control: tags -1 upstream
Control: retitle -1 Upstream testsuite fails to produce deterministic results

Santiago,

On Sun, Oct 29, 2023 at 02:39:44PM +0100, Santiago Vila wrote:
> However, I can create a machine for you to reproduce the problem.

Thanks. I have reproduced on your provided machine, but still not locally and I
can't identify the underlying difference between the builds.

In the failure case the problem is in the upstream testsuite, specifically the
test for #491391 in tests/run-testsuite which produces

init.d:
bootchart
four
one
rmnologin
three
two

insserv:
override

rc0.d:

rc1.d:

rc2.d:
S01one
S01three
S01two
S02four
S98rmnologin
S99bootchart

rc3.d:
S01one
S01three
S01two
S02four
S98rmnologin
S99bootchart

rc4.d:
S01one
S01three
S01two
S02four
S98rmnologin
S99bootchart

rc5.d:
S01one
S01three
S01two
S02four
S98rmnologin
S99bootchart

rc6.d:

rcS.d:
error: incorrect 5 sequence bootchart not before rmnologin

The same failure mode appears to be responsible for armel and armhf autopkgtest
failures logged on ci.debian.net[1]

As Ian pointed out[2], there are significant and surprising changes in looping
order and behaviour between the successful and failing testsuites. The diff is attached.

Having said that, I still can't reproduce locally or determine a good fix.
Hopefully Jesse will have a useful contribution

Mark

[1]  https://ci.debian.net/data/autopkgtest/unstable/armel/i/insserv/38435862/log.gz

[2]  https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1052942#15

-------------- next part --------------
A non-text attachment was scrubbed...
Name: insserv.build.diff
Type: text/x-diff
Size: 52640 bytes
Desc: not available
URL: <http://www.chiark.greenend.org.uk/pipermail/debian-init-diversity/attachments/20231031/8dd72694/attachment-0001.diff>