-
Notifications
You must be signed in to change notification settings - Fork 150
Segfault on linux #121
Comments
Happens with bcolz |
@FrancescAlted |
That's possible yes. I would say that downgrading to 1.4.x would be a sensible action until 1.5 is fixed. Enviat des del meu iPhone El 15/01/2015, a les 23:37, Valentin Haenel notifications@github.com va escriure:
|
This is a bit crazy because I cannot reproduce the problem on my Ubuntu 14.10 machine. With Ubuntu system Python:
With Anaconda Python:
Hmm, tricky... |
I continue to have the feeling that perhaps some recent optimization in the default blosclz has something to do with the problem. Could you please use the next branch (basically master but using blosclz 1.0.2): https://github.com/Blosc/c-blosc/tree/blosclz-1.0.2 and tell me how it goes with your machine? |
In case the above works, and to make sure that the problem is in blosclz 1.0.3 I have created another branch using blosc 1.4.1 but with blosclz 1.0.3 (instead of 1.0.2): https://github.com/Blosc/c-blosc/tree/blosc-1.4.1-blosclz-1.0.3 Please check it out and tell me how it goes. This can also be done with #120 and #122. |
#166 by itself doesn't fix this with blosc |
#166 + blosc master at Blosc/c-blosc@8250362 doesn't fix it either |
Hmm #166 + https://github.com/Blosc/c-blosc/tree/blosc-1.4.1-blosclz-1.0.3 seems to be fine, |
But #166 with https://github.com/Blosc/c-blosc/tree/blosclz-1.0.2 doesn't |
Well, that excludes blosclz 1.0.3 for being guilty, so that's something. Uh, I wonder I can't reproduce the crash on my laptop. Would that be a problem with the compiler or pthreads implementation? Here are my versions:
Can you paste yours? |
This no longer happends with 2.7.10 from anaconda and current master at e23e79f |
I am using Debian stable:
And I am seeing segfaults, only when running bcolz tests using the builtin test runner and only outside my home directory:
This was built using a default build of Python from source. It seems like there is something wrong right before exit, as the backtrace confirms:
Then, this also is an issue on an Anaconda Python:
And the backtrace:
Works fine with nosetests on both Pythons.
Amazingly enough, it also works with Blosc 1.4.1:
And with anaconda python:
The text was updated successfully, but these errors were encountered: