Bug#906027: shellcheck: unable to decommit memory: Invalid argument

[Helmut Grohne]

Control: forwarded -1 https://github.com/koalaman/shellcheck/issues/785

On Mon, Aug 13, 2018 at 12:15:41PM +0000, Luke Goodsell wrote:
> In Debian 9, running shellcheck against a moderate-size file results in the error message
> 
>     shellcheck: unable to decommit memory: Invalid argument
> 
> I can reproduce this with these commands:
> 
>     docker run -it debian:9
>     apt-get -y update
>     apt-get -y install shellcheck
>     echo '#!/bin/bash' > test.sh; for i in $(seq 1000); do echo 'echo "hi"' >> test.sh; done
>     shellcheck test.sh
>     # shellcheck: unable to decommit memory: Invalid argument
>     # shellcheck: unable to decommit memory: Invalid argument
>     # shellcheck: unable to decommit memory: Invalid argument
> 
> Since shellcheck is used in CI tests and produces no output on success, these error messages shouldn't be present.
> 
> Uname: Linux 0b9a6fddbfc7 3.10.0-862.3.3.el7.x86_64 #1 SMP Fri Jun 15 04:15:27 UTC 2018 x86_64 GNU/Linux
> Libc6: Version: 2.24-11+deb9u3

Thank you very much for the report and the precise version information.

Researching the problem, I see two major ways forward.

The issue is reported against the upstream tracker in github and is
being discussed there. As of yet, there is no solution known.

The other is that the message is not quite shellcheck-specific but more
related to the underlying tooling (ghc and maybe glibc). Joey Hess
experienced a very similar issue with git-annex, see #847677. It got
fixed in ghc/8.0.1-15. Now shellcheck 0.4.4-4 amd64 (the one from the
Debian archive) was built with ghc 8.0.1-8. Possibly rebuilding the
package with a newer ghc (and the one from stable suffices) could make
it go away.

Given that I am not reproducing the issue for myself on stretch, can I
ask you to perform the following tests?

 * Try debootstrapping a buster or sid system on the same kernel
   version. Install shellcheck from buster or sid. Try running it and
   observe whether it fails in the same way.

 * Rebuild shellcheck version 0.4.4-4 (the one from stable) in a stable
   environment with no source changes. Does that rebuilt shellcheck
   package also reproduce the issue?

>From my present understanding of the issue, I expect that both tests
make the issue disappear. If that is indeed the case, the issue is fixed
in buster and sid and it can be fixed in stable by performing a binNMU.
In that case, you can send the following to control at bugs.debian.org:

reassign 906027 release.debian.org
retitle 906027 nmu: shellcheck_0.4.4-4
user release.debian.org at packages.debian.org
usertags 906027 + binnmu

The release team would want to know:

nmu shellcheck_0.4.4-4 . ANY . stretch . -m "rebuild against ghc with #847677"

Thanks for your assistance

Helmut