#974004 perl: FTBFS on m68k and sh4: qemu-user trashes argv[0] breaking multi-call binaries

Package:
src:perl
Source:
perl
Submitter:
Dominic Hargreaves
Date:
2024-11-24 19:21:01 UTC
Severity:
important
Tags:
#974004#5
Date:
2020-11-08 21:48:53 UTC
From:
To:
We seem to be affected by #970460 in qemu-user which is breaking builds
on m68 and sh4:

https://lists.debian.org/debian-perl/2020/11/msg00007.html

Severity set to important since these are non-release architectures.

#974004#12
Date:
2021-02-02 13:35:54 UTC
From:
To:
This is an ongoing issue that occurred for 5.32.1 too. For future
new upstream releases, we could send a courtesy note along the lines
of <https://alioth-lists.debian.net/pipermail/perl-maintainers/2021-February/007086.html> (ideally just before we upload).

#974004#19
Date:
2024-11-23 18:31:03 UTC
From:
To:
Hi!

The issue in qemu has been fixed a LONG time ago, it retains argv[0]
since before buster.

What's up here?

/mjt

#974004#24
Date:
2024-11-24 09:20:34 UTC
From:
To:
I don't know really. All I know is there's buildd logs with the same
failure mode as recently as 2024-09 on sh4 and 2023-11 on m68k.

https://buildd.debian.org/status/fetch.php?pkg=perl&arch=sh4&ver=5.40.0-4&stamp=1725235748&raw=0

https://buildd.debian.org/status/fetch.php?pkg=perl&arch=m68k&ver=5.38.0-2&stamp=1701215410&raw=0

But I've pretty much stopped caring about build results on those
architectures, and haven't sent any courtesy notes for the last three
transitions. Adrian (who originally requested heads-up notes to porters
back in 2020) hasn't complained, so I guess things work well enough.

Closing this but copying Adrian fyi.

#974004#25
Date:
2024-11-24 16:42:38 UTC
From:
To:
Hello Niko,

This issue is not specific to m68k and sh4, but it's an issue with qemu-user.

You will observe this on any architecture being emulated with the help of
qemu-user, including riscv64 or arm64.

But why close it when it still sometimes shows up? Better re-assign it to
src:qemu. Bug reports are supposed to document bugs or issues, it's better
for users to keep them publicly visible as they provide information for
anyone running into this issue as well.

I suggest re-opening and re-assigning the bug to src:qemu.

Adrian

#974004#30
Date:
2024-11-24 17:49:23 UTC
From:
To:
24.11.2024 19:42, John Paul Adrian Glaubitz wrote:

It's been fixed in qemu before buster release.  The qemu bug is #970460.

If there are issues with it still, I don't know about it, and I want to
hear it, pretty please.

/mjt

#974004#35
Date:
2024-11-24 19:17:18 UTC
From:
To:
Control: reopen -1

I'm aware.

And it's supposed to be long fixed in qemu, so I assume (some part of)
the fix is just still not deployed on some of the experimental buildds
for m68k and sh4. Or whatever. In any case, I don't recall a single
buildd failure with this failure mode on other architectures.

I don't really mind reopening, but this is not going to be fixed by
any change to src:perl. We were using the bug as a reminder about your
request to send advance warnings about Perl transitions because of the
issue. We stopped those as they didn't seem useful.

Feel free to reassign if you think there's still something on the
qemu side to fix. I lack the visibility (and tbh the inclination) to
investigate this further myself.