New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Build-bot status #656

Closed
arokem opened this Issue May 25, 2015 · 15 comments

Comments

Projects
None yet
3 participants
@arokem
Member

arokem commented May 25, 2015

This can be seen on the buildbots:

http://nipy.bic.berkeley.edu/builders/dipy-py2.7-osx-10.8/builds/362/steps/shell_5/logs/stdio

And I also get this on my machine after pulling from master. I believe @RafaelNH is seeing something similar on his machine as well.

@omarocegueda

This comment has been minimized.

Contributor

omarocegueda commented May 25, 2015

It seems like it is complaining about cimport-ing inline functions. I think it makes sense that doing that is not valid, but it's weird that it works on other platforms (e.g. Travis and windows)

@arokem

This comment has been minimized.

Member

arokem commented May 26, 2015

It's failing on Windows as well:
http://nipy.bic.berkeley.edu/builders/dipy-bdist32-33/builds/126/steps/shell_6/logs/stdio

On Mon, May 25, 2015 at 4:58 PM, Omar Ocegueda notifications@github.com
wrote:

It seems like it is complaining about cimport-ing inline functions. I
think it makes sense that doing that is not valid, but it's weird that it
works on other platforms (e.g. Travis and windows)


Reply to this email directly or view it on GitHub
#656 (comment).

@omarocegueda

This comment has been minimized.

Contributor

omarocegueda commented May 26, 2015

Oh! you're right! I tested this on linux and windows 8.1 with no issues. I'm checking what's the problem with defining pointers to inline functions (that's what cython is trying to do). It may be an undefined behavior which is handled differently by different compilers.

@omarocegueda

This comment has been minimized.

Contributor

omarocegueda commented May 26, 2015

I think this is a cython bug. We can actually define a pointer to an inline function as follows:

#include <stdio.h>
int inline my_inline(int x){
    return 2*x;
}
static int (*fpointer)(int);

int main(){
    fpointer = my_inline;
    int y = fpointer(5);
    printf("%d\n", y);
    return 0;
}

The above code compiles and runs with output 10. However if we add inline to the fpointer variable, it will of course fail, because a variable cannot be declared as inline. So, Cython should remove the CYTHON_INLINE prefix to the declaration. However we should see this error everywhere, not just on some platforms... still investigating...

@omarocegueda

This comment has been minimized.

Contributor

omarocegueda commented May 26, 2015

Ok, according to his:
http://docs.cython.org/src/tutorial/pxd_files.html
inline functions should be defined (not just declared) in the pxd. I'll try_branch with that...

@omarocegueda

This comment has been minimized.

Contributor

omarocegueda commented May 26, 2015

Oh! no! I no longer have access to the buildbots =(
WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED!
I cannot try_branch, any ideas?

@matthew-brett

This comment has been minimized.

Member

matthew-brett commented May 26, 2015

@arokem

This comment has been minimized.

Member

arokem commented May 26, 2015

#657 seems to have helped with a lot of our bots, but this one seems to be lacking nibabel:

And there's this mysterious one:

Other bots are now up-to-date with recent scipy, so they have this one:

That should be fixed with #612

This one's new (to me), and most likely related to #629:

Anyone know what that's all about?

@matthew-brett

This comment has been minimized.

Member

matthew-brett commented May 26, 2015

I think I fixed the problem with nibabel being missing - that was a configuration change for the buildbots, that I hadn't finished cleaning up.

I can replicate:

http://nipy.bic.berkeley.edu/builders/dipy-py3.3/builds/369/steps/shell_6/logs/stdio

with just np.load('dev_trees/dipy/dipy/data/t1_coronal_slice.npy') on Python 3.3 using numpy 1.7.1 through 1.8.1 (haven't tested 1.8.2 or 1.9.0). Maybe this is a problem with the npy file format?

@matthew-brett

This comment has been minimized.

Member

matthew-brett commented May 26, 2015

Relevant SO post here: http://stackoverflow.com/questions/27145273/compatibility-numpy-1-7-vs-numpy-1-8

I just tried loading and resaving the file with numpy 1.9.2 - seems to work.

#658

@matthew-brett

This comment has been minimized.

Member

matthew-brett commented May 26, 2015

Just set off a build : http://nipy.bic.berkeley.edu/builders/dipy-py3.3/builds/370

I won't be online to see it finish, probably.

@arokem

This comment has been minimized.

Member

arokem commented May 26, 2015

OK - I will monitor that and merge if/when it comes back green.

On Tue, May 26, 2015 at 2:54 PM, Matthew Brett notifications@github.com
wrote:

Just set off a build :
http://nipy.bic.berkeley.edu/builders/dipy-py3.3/builds/370

I won't be online to see it finish, probably.


Reply to this email directly or view it on GitHub
#656 (comment).

@matthew-brett

This comment has been minimized.

Member

matthew-brett commented May 26, 2015

Looks to be passing the tests that were failing earlier :

http://nipy.bic.berkeley.edu/builders/dipy-py3.3/builds/370/steps/shell_6/logs/stdio

@arokem arokem changed the title from Compilation failures on master to Build-bot status May 26, 2015

@omarocegueda

This comment has been minimized.

Contributor

omarocegueda commented May 27, 2015

Thanks Matthew!,
I guess it should be possible to detect when a function is inline and just
remove the CYTHON_INLINE from the pointer declaration in that case. But
since moving the inline functions to the pxd solved the issue (and since it
is actually documented), it would be more a feature request rather than a
bug.
EDIT: I just sent an email to the Cython mailing list explaining the issue.
On Tue, May 26, 2015 at 9:39 AM, Matthew Brett notifications@github.com
wrote:

Omar - sorry - we comprehensively upgraded the server, and that included
getting a new server key.

It's no problem I think - however you are using ssh, you should get some
more text with that error telling you that, if you believe the server is
OK, you should delete the old line from the known_hosts file and try
again.

For the Cython bug - if you think this is a bug - would you mind reporting
to the Cython issues or mailing list?


Reply to this email directly or view it on GitHub
#656 (comment).

"Cada quien es dueño de lo que calla y esclavo de lo que dice"
-Proverbio chino.
"We all are owners of what we keep silent and slaves of what we say"
-Chinese proverb.

http://www.cimat.mx/~omar

@arokem

This comment has been minimized.

Member

arokem commented Sep 1, 2015

This one seems to be outdated - closing.

@arokem arokem closed this Sep 1, 2015

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment