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

Trying to be smarter than autotools #1

Open
ahmadia opened this issue Oct 8, 2013 · 6 comments
Open

Trying to be smarter than autotools #1

ahmadia opened this issue Oct 8, 2013 · 6 comments

Comments

@ahmadia
Copy link
Contributor

ahmadia commented Oct 8, 2013

Looking at the code in autotools_package.py, I see explicit hardcoding with a hook function for rpath support on Linux.

In general, I would be more comfortable if our default behavior was to trust the build system (in this case, modern autotools should do a pretty good job of building the right thing on both Linux and OS X), instead of trying to be smarter than it. I'm really not interested in reinventing all the work that goes into autotools, CMake, etc...

Thoughts from others?

@dagss
Copy link
Member

dagss commented Oct 8, 2013

If you can find the flag to make autotools build libraries with RPATH linkage, then please go ahead and fix it.

@ahmadia
Copy link
Contributor Author

ahmadia commented Oct 8, 2013

It should be automatically added on systems where it's supported. I'm guessing this wasn't the case for you?

@dagss
Copy link
Member

dagss commented Oct 8, 2013

It has been the case at one point or another, though I haven't tried recently (and perhaps it depends on whether you pass --with-foo or whether the library is auto-detected, the flags basically allows auto-detection to pick it up without any --with-foo. E.g., HDF doesn't give an option for zlib, and so on.).

I suggest changing the default of the set_FLAGS (or whatever it was) to False, and then turn it on to true whenever it is needed.

@ahmadia
Copy link
Contributor Author

ahmadia commented Oct 8, 2013

okay, that seems like a reasonable compromise :)

I think we may also need to move the rpath function somewhere else, since
it's conceivable that other builder scripts may want/need access to it.

On Tue, Oct 8, 2013 at 12:21 PM, Dag Sverre Seljebotn <
[email protected]> wrote:

It has been the case at one point or another, though I haven't tried
recently (and perhaps it depends on whether you pass --with-foo or
whether the library is auto-detected, the flags basically allows
auto-detection to pick it up without any --with-foo. E.g., HDF doesn't
give an option for zlib, and so on.).

I suggest changing the default of the set_FLAGS (or whatever it was) to
False, and then turn it on to true whenever it is needed.


Reply to this email directly or view it on GitHubhttps://github.com//issues/1#issuecomment-25904979
.

@dagss
Copy link
Member

dagss commented Oct 8, 2013

Yes, that was the intention, just never got around to it. There's support to specify hook_import_dirs or similar in the profiles, just set it to base (or create a new directory lib or something), and then all the hooks can import modules placed there.

@certik
Copy link
Member

certik commented Oct 8, 2013

I don't think I ever managed to get rpath working with autotools without forcing it.

cekees added a commit that referenced this issue Apr 2, 2014
Update to latest hashstack master, add Jasper host package on Cygwin
lanewittgen added a commit that referenced this issue Apr 14, 2014
add verify functionality to matplotlib build
johannesring added a commit that referenced this issue Jul 1, 2016
set MAX_JBUFS=24 in parmetis on OS X
cekees added a commit that referenced this issue Dec 10, 2017
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

No branches or pull requests

3 participants