Re: [PATCH v2 glibc 0/2] SIGSTOP/SIGCONT duplicating portions of files.

Hi Samuel,

On 28/06/2026 18:16, Samuel Thibault wrote:
> Michael Kelly, le ven. 26 juin 2026 20:11:02 +0100, a ecrit:
>> Is it expected that the corruption of a few months ago would still occur ?
> I don't know.
>
> But for instance, haskell-pandoc still fails to build:
>
> <no location info>: error:
>      /usr/lib/haskell-packages/ghc/lib/x86_64-hurd-ghc-9.10.3-inplace/libHSserialise-0.2.6.1-LnGm3cCB6Id4o0hFZmypWQ-ghc9.10.3.so: undefined symbol: cborgzm0zi2zi10zi0zm3W6EjwNwxXuID7SvuFmtY9_CodecziCBORziEncoding_TkFloat64_con_info
>
> While that symbol does exist in libghc-cborg-dev's
> libHScborg-0.2.10.0-3W6EjwNwxXuID7SvuFmtY9-ghc9.10.3.so, which is pulled
> by libghc-serialise-dev

I rebuilt libghc-cborg-dev locally and comparing the hexdump of the 
resulting shared library with that from version 'haskell-cborg 
0.2.10.0-3+b2' shows the same pattern of corruption that we have seen 
before. There are several batches of 16 bytes which are non-zero in my 
locally built library but replaced with zeroes at offsets 0x50, 0x70 and 
0xff0 in the 'sid' package. I ran your 'checkone' script (using readelf 
and ld to test library validity) on both of the associated .deb packages 
but it doesn't find any errors in either.

I think the build log for the 'sid' package is:

https://buildd.debian.org/status/fetch.php?pkg=haskell-cborg&arch=hurd-amd64&ver=0.2.10.0-3%2Bb2&stamp=1775562754&raw=0

This log shows it was built against libc0.3-dev_2.42-14~hurd.1 which 
wouldn't include the SIGSTOP/SIGCONT fix. There was a thought that there 
has been less corruption since that fix was made, so I'm wondering if it 
would be helpful to rebuild and release any of the haskell packages that 
were built using libc0.3 earlier than 2.42-15.

Is it feasible to identify these packages ?

Regards,

Mike.