Re: [Linuxsampler-devel] [PATCH] Fix endianness and memory alignment issues for big-endian architectures (s390x)

[Christian S. <sch...@li...>]

On Friday, 14 August 2026 10:02:48 CEST Christian Schoenebeck wrote:
> On Thursday, 13 August 2026 15:38:38 CEST Christian Schoenebeck wrote:
> > On Tuesday, 11 August 2026 11:19:03 CEST Pranav P wrote:
[...]
> > Hi Pranav,
> > 
> > thanks for pointing this out and for the patch.
> > 
> > I made a quick test with QEMU s390x user mode emulation and was able to
> > reproduce these BE issues.
> > 
> > Give me some time. I am thinking about a cleaner approach to fix these.
> 
> I have split your fixes up into separate commits.
> 
> For the following fixes I chose a different, cleaner approach:
> 
> https://svn.linuxsampler.org/cgi-bin/viewvc.cgi?view=revision&revision=4651
> https://svn.linuxsampler.org/cgi-bin/viewvc.cgi?view=revision&revision=4652
> 
> For the following fix I picked your solution, just brought back comments
> that your dropped:
> 
> https://svn.linuxsampler.org/cgi-bin/viewvc.cgi?view=revision&revision=4653
> 
> All existing tests already pass for s390x here for me. However there is
> still one fix in your patch that I have not addressed in SVN yet: your BE
> changes for calling __calculateCRC(). How did you hit this issue, given
> that all tests already succeed for me with current SVN r4653?

And the last issue from your patch is now fixed with different approach by:

https://svn.linuxsampler.org/cgi-bin/viewvc.cgi?view=revision&revision=4656
https://svn.linuxsampler.org/cgi-bin/viewvc.cgi?view=revision&revision=4660

I guarded this BE issue with a new test case that bypasses the libgig API to 
avoid the bug being masked, which was the reason why the tests passed even 
without this fix applied.

I consider your reported BE issues being fixed now. Otherwise let me know.

Thanks!

/Christian