8000 gfortran: error: unrecognized command line option ‘-ffpe-summary=invalid,zero’ · Issue #6495 · numpy/numpy · GitHub
[go: up one dir, main page]

Skip to content

gfortran: error: unrecognized command line option ‘-ffpe-summary=invalid,zero’ #6495

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

Closed
pv opened this issue Oct 17, 2015 · 12 comments
Closed

Comments

@pv
Copy link
Member
pv commented Oct 17, 2015

numpy.distutils adds a -ffpe-summary command line option to gfortran, but it's apparently not supported by all versions of the compiler.

See scipy/scipy#5380

Apparently this was added in b354c00 --- seems to affect also non-windows platforms.

@charris
Copy link
Member
charris commented Oct 17, 2015

@carlkl Comment?

@pv
Copy link
Member Author
pv commented Oct 17, 2015

The part where these are added probably should be guarded with is_win64().
-mlong-double-64 and -DMS_WIN64 probably should also only be added for mingw (and not e.g. for unix)

@pv
Copy link
Member Author
pv commented Oct 17, 2015

The mingw64 changes seem to have also changed some other flags on non-Windows platforms --- these should probably also be re-reviewed changed back just to be sure.

@charris
Copy link
Member
charris commented Oct 17, 2015

@pv I could also revert the PR. I didn't think it quite ready but put it in as a gamble to see what happened. The *.*.1 releases are the FU releases ;)

@pv
Copy link
Member Author
pv commented Oct 17, 2015

Note that also gh-5614 seems to add some win64 stuff to all platforms.

It seems lots of the stuff should be guarded by if is_mingw64py(): ......

@carlkl
Copy link
Member
carlkl commented Oct 17, 2015

it seems I forgot to put this flags in gnu.py with in a elif is_win64(): clause. Too bad.
Another problem I encountered is, that is is necessary to discriminate mingw-w64|mingwpy and the old mingw32 based numpy-vendor compiler as longdouble == double should not happen there.

@charris charris removed this from the 1.10.2 release milestone Oct 17, 2015
@charris
Copy link
Member
charris commented Oct 20, 2015

See also #6473. @carlkl I'm tempted to revert both #5614 (#5587) and #6354.

@carlkl
Copy link
Member
carlkl commented Oct 20, 2015

@charris, maybe it is better to redo a clean mingwpy PR from the scratch. So feel free to revert all these PR's.

@charris charris added this to the 1.10.2 release milestone Oct 20, 2015
@charris charris mentioned this issue Oct 20, 2015
@carlkl
Copy link
Member
carlkl commented Oct 21, 2015

@charris, howto do a new PR now? Against https://github.com/numpy/numpy/tree/maintenance/1.10.x ?

@charris
Copy link
Member
charris commented Oct 23, 2015

@carlkl All PRs start in master. Numpy 1.11 should be coming out in about 4 months, so that is probably a good release to shoot for in getting OpenBLAS support up and running.

@charris
Copy link
Member
charris commented Oct 23, 2015

@pv This should be fixed now.

@charris
Copy link
Member
charris commented Oct 25, 2015

OK, closing this. We should be back to the 1.9 state which, AFAICT, was not causing problems.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants
0