-
-
Notifications
You must be signed in to change notification settings - Fork 452
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
configure.ac: write build/make/Makefile within an AC_CONFIG_FILE, not during main configure #21524
Comments
The title "Get rid of file descriptor 7 hack" is different from the description "build/make/Makefile should be written within an AC_CONFIG_COMMANDS procedure". Do you want to fix both things here? Can you please clarify? And honestly, I think there is nothing wrong with the "file descriptor 7 hack". |
comment:2
I've updated the title |
comment:3
I agree, there's nothing wrong with using a file descriptor to write to a file ;) |
This comment has been minimized.
This comment has been minimized.
Dependencies: #21479 |
This comment has been minimized.
This comment has been minimized.
comment:6
Does it necessarily need to use As I see it the main goal here is to simplify the configure.ac by moving more functionality into separate files. But moving the existing code into an m4 macro would take less work, and has the advantage that it would have full access to any other variables set in other parts of the |
comment:7
On second thought, I can see why this way would be better, as it's more in spirit with how configure is supposed to work w.r.t. generating files. I'll have to give that some more thought. Maybe there aren't so many variables that need to be passed in. |
Branch: u/embray/build/makefile-in |
This comment has been minimized.
This comment has been minimized.
Commit: |
Author: Erik Bray |
comment:8
Here's my proposed solution to this (I updated the ticket to mention This adds a If any of this should be controversial, it may be the use of some more advanced GNU make constructs, the portability of which I'm not sure. This was tested with GNU make 3.81 (about 11 years old), and 4.2.1 (very recent). That said, it's using some relatively advanced GNU make features that may not be available in other makes. I think it's still better, but if need be more of the Makefile generation could be pushed back into Increased the priority since I have other build system work that would likely depend on this, and also because it makes some non-trivial performance improvements, especially on Cygwin. New commits:
|
comment:10
Weird that the patchbot is failing due to Numpy of all things. I wonder if it has something more to do with the fact that it's building in a temp dir (since this is an "unsafe" ticket) than anything to do with the patch itself, but I'll have to see if I can reproduce...
|
Work Issues: merge conflicts |
comment:58
Ok. Your recent changes look good to me. |
Changed branch from u/embray/build/makefile-in to |
Changed commit from |
comment:60
This completely breaks upgrading Sage, it seems to me. On three different OS X machines with 8.2.beta7 installed, upgrading to 8.2.beta8, or in particular just using this branch, results in a bad
Please fix this. |
comment:61
Indeed, it looks like a bug introduced upon the last rebase. I didn't catch it because it only affects if you are using Sage's |
comment:62
Actually it appears the problem might come from #24703 in the first place, since it seems to misplace the space in the |
comment:65
You cannot just re-open a ticket. Only the release manager should do that. |
comment:66
Especially this ticket must not be re-opened since it has already been released. Just open a new ticket to fix it. |
comment:67
It's not in the latest beta yet. |
comment:68
Oops, apparently it is. If I had thought otherwise I wouldn't have reopened. Strange--I just checked too any I didn't see it. |
comment:69
Followup in #24961. |
comment:70
Thanks for the quick diagnosis, by the way! |
comment:71
Replying to @jhpalmieri:
No, thank you for pointing it out. |
comment:72
Why is this making changes to |
comment:73
This might have broken |
comment:74
Replying to @jdemeyer:
I just noticed this comment. That was not intentional. Probably a mis-resolved conflict (though I don't know why since I never made any edits to that file in the first place). |
comment:75
|
comment:76
What was the purpose of this change: @@ -58,11 +58,11 @@ all-sage: \
# option to make forces all targets to be built unconditionally)
download-for-sdist: base
env SAGE_INSTALL_FETCH_ONLY=yes $(MAKE) -B SAGERUNTIME= \
- $(SDIST_PACKAGES)
+ $(SDIST_PACKAGE_INSTS)
It has broken sdists, see #25319. |
build/make/Makefile
should be written byconfig.status
within anAC_CONFIG_FILE
template, not atconfigure
time.Depends on #24729
Depends on #24703
Component: build: configure
Author: Erik Bray
Branch:
2a99185
Reviewer: Julian Rüth
Issue created by migration from https://trac.sagemath.org/ticket/21524
The text was updated successfully, but these errors were encountered: