[Debian-med-packaging] Bug#976929: parsinsert: FTBFS on ppc64el: incorrect result

[Étienne Mollier]

Control: tags -1 pending

Hi David,

Following up on parsinsert failing to build from source on POWER
64 bits little endian on Debian Testing, I found out that
reducing the compiler optimization level to -O1 allows the test
suite to pass.

The affected configuration is Debian 11 at Testing stage, with
Gcc 10.2.  Affected CPU architectures are armel, armhf,
mips64el, ppc64el (which is the subject of Debian Bug#976929[1])
and riscv64.

Normally, the package is built for Debian with optimization
level -O2, and this passes for all other architectures,
including but not limited to amd64, arm64, i386, mipsel, and
s390x.  Interestingly, when pushing to -O3, the problem may
appear on amd64 architecture, see #964082[2], but it does not
seem to be reproducible on every amd64 CPU though.  In any case,
the erroneous results in output are the same as on more exotic
architectures.

[1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=976929
[2] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=964082

I also gave a shot to -Os, sitting in between -O1 and -O2, and
most interestingly, I had a rather different output:

	[...]
	Loading sequence 180000
	Loading sequence 200000
	Loading sequence 220000
	 taxomony index file: 224931 taxa
	Reading Newick Tree: [core_rdp.ntree]
	opening source file: [core_rdp.ntree]
	Opening file: core_rdp.ntree    ...Parsing Newick file: core_rdp.ntree  ...Processed 246406 chars of 246447
	Newick Tree Read Completed: 10143 taxa, 0 sec
	Found duplicate name: [W<85><9c>Ý<9c>U]/nFound duplicate name: [W<85><9c>Ý<9c>U]/n...
	Loading sequence 10143 [51224]
	ERROR - tree member sequence is not available [W<85><9c>Ý<9c>U]
	ERROR - tree member sequence is not available [W<85><9c>Ý<9c>U]
	[...]
	ERROR - tree member sequence is not available [W<85><9c>Ý<9c>U]
	ERROR - tree member sequence is not available [W<85><9c>Ý<9c>U]
	Process Completed: 1 sec
	incorrect result

Note I had non printable characters between the bracket, so may
be pointing to uninitialized memory, or similar.  Not sure how
relevant this is though.

In hope this helps,
-- 
Étienne Mollier <etienne.mollier at mailoo.org>
Fingerprint:  8f91 b227 c7d6 f2b1 948c  8236 793c f67e 8f0d 11da
Sent from /dev/pts/5, please excuse my verbosity.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/debian-med-packaging/attachments/20210119/0909398a/attachment-0001.sig>