Skip to content
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

FIX: Fixes the msvcr problem in MinGW #363

Closed
wants to merge 1 commit into from

Conversation

certik
Copy link
Contributor

@certik certik commented Jul 20, 2012

See this thread for more information:

http://mail.scipy.org/pipermail/numpy-discussion/2012-July/063381.html

I think this should go in before the release, because otherwise the sources don't build out of the box using MinGW. The 1.6.2 release builds just fine using the same Windows setup.

/cc @teoliphant, @charris, @rgommers

@travisbot
Copy link

This pull request fails (merged 76b1faf into f93774d).

@cournape
Copy link
Member

I -1 on this as is. It is just removing code that does not work but used to. It may be an actual bug, but we don't know what it is.

@certik
Copy link
Contributor Author

certik commented Jul 20, 2012

In the mailinglist thread (see above for a link), the bug seems to be that objdump can't read the msvcr.dll. What is the rationale to build such a library on our own? I think we need to do something, at least add a better error message, but I thought the suggestion on the mailinglist was to simply remove the code.

@cournape cournape closed this Jul 20, 2012
@cournape
Copy link
Member

I cannot reproduce the issue, and confirmed that this change breaks the build. This fix is wrong.

@certik
Copy link
Contributor Author

certik commented Jul 22, 2012

@cournape, thanks a lot for your time and the review.

Can you please post the log of the break? It looks like you have some different setup that I cannot reproduce. Which mingw version do you use? Do you use Wine or Windows (which one)?

Also, given your experience, what do you think would be a better fix?

@certik
Copy link
Contributor Author

certik commented Jul 22, 2012

I must have misunderstood Nathan's suggestion to just remove the code: http://mail.scipy.org/pipermail/numpy-discussion/2012-July/063390.html

@cournape
Copy link
Member

Ondrej: I used windows 7 64 bits (but use 32 bits python 2.7.3 and 32 bits toolchain). As for MinGW, I use the one given by the 5.0.4 installer referred by Ralf.

I don't think Nathan's is right, in the sense that the msvc runtime 9 import library is required whether we use localtime_s or not. Python 2.6 (and later, up to 3.2), require this version of the runtime, which is not included in the mingw version I used.

Unfortunately, using a more recent version for MinGW (using gcc 4.x) is not very practical for an installer, because it requires the deployment of additional dlls. Deploying shared dlls on windows is a big PITA: the only one I know is to add the dll in the python directory, which is very ugly. I don't know of any other solution for windows XP/Vista (windows 2008 and 7 have a way to support this).

luyahan pushed a commit to plctlab/numpy that referenced this pull request Apr 25, 2024
feat: Add vfma[s|d]_laneq_[f32|f64]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants