-
Notifications
You must be signed in to change notification settings - Fork 271
Testsuite fails on several archs #77
Comments
Hello, there seem to be some long-standing problems with binary data. I am rewriting the PocketSphinx build system, would it be possible for you to test my fork at https://github.com/dhdaines/pocketsphinx ? Sadly I have not been a Debian developer for several decades so do not have access to all these wonderful platforms anymore :) |
This is still failing to a large extent, e.g. on s390x:
|
Hmm, good to know - big-endian platform I assume? Can you rerun with |
also the contents of build/config.h would be helpful! |
yes, it's big-endian, here they are: |
Thanks! It seems that a new binary language model format was implemented sometime after PocketSphinx 0.8 with zero consideration for byte ordering. Currently considering if this is another 5prealpha "feature" that should just get reverted... |
I'm guessing the Alpha problem has something to do with unaligned access to memory. In which case it is probably my fault :) |
Most of the tests should be fixed on big-endian machines by https://github.com/dhdaines/pocketsphinx/commit/192ac3f1141558d91381bc1700bddbabed4d878b - please try and let me know! |
Hello,
The sphinxbase testsuite fails on some architectures in Debian :
ppc64el: https://buildd.debian.org/status/fetch.php?pkg=sphinxbase&arch=ppc64el&ver=0.8%2B5prealpha%2B1-8&stamp=1582414870&raw=0
alpha: https://buildd.debian.org/status/fetch.php?pkg=sphinxbase&arch=alpha&ver=0.8%2B5prealpha%2B1-8&stamp=1583125429&raw=0
Samuel
The text was updated successfully, but these errors were encountered: