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

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

On Thursday, 20 August 2026 09:29:47 CEST Pranav P wrote:
> Hi Christian,
> 
> Sorry for the delayed response. I tested out your patch and it does fix the
> issue. Thanks a lot for pitching in.

In the meantime I have committed additional BE fixes to SVN, e.g. for 
SoundFont. Some more might still follow. I am confident that next maintenance 
release will have them covered.

https://svn.linuxsampler.org/cgi-bin/viewvc.cgi?view=revision&revision=4662
https://svn.linuxsampler.org/cgi-bin/viewvc.cgi?view=revision&revision=4663


> So, the reason that I made the final changes (the ones even after the test
> cases passed) was because I have seen circumstances where issues might be
> present in both reading and writing which gets masked away. I tested them
> by generating files (a.riff, b.riff, c.riff, d.riff and foo.gig) in the
> src/testcases folder in a little-endian machine, and copying them into a
> big-endian machine and then running the test-suite after commenting out the
> test cases which write into them (so that only the reading part is tested).

Ah, that makes sense then. Because it would only manifest exactly if the RIFF 
files were written by LE machine and read by BE machine, so I already assumed 
you did this manually.

I have added test cases that bypass the libgig API and preventing that double-
wrong on the same endian-type host from masking the tests pass on broken 
endian code, e.g. with r4662 and:

https://svn.linuxsampler.org/cgi-bin/viewvc.cgi?view=revision&revision=4664

Thanks!

/Christian