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
Upgrade cddlib to 0.94j #25344
Comments
comment:2
Why wait? If we have the tarball we can do it now, and it will be uploaded to Sage's mirrors. |
comment:3
Because I would still like to get your cygwin fixes into upstream. |
comment:4
Ah, ok. |
Branch: u/saraedum/25344 |
Commit: |
Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:
|
Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:
|
Work Issues: do all doctests pass? |
comment:8
I am running --long doctests for all of Sage currently; everything in |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Changed work issues from do all doctests pass? to do all doctests pass?, the tarball is not final yet |
Changed work issues from do all doctests pass?, the tarball is not final yet to the tarball is not final yet |
This comment has been minimized.
This comment has been minimized.
comment:15
ccing people who previously touched the cddlib integration code. |
Changed work issues from the tarball is not final yet to do all doctests pass? |
Branch pushed to git repo; I updated commit sha1. New commits:
|
New commits:
|
comment:20
I assume this has the Cygwin fixes in it now? |
comment:27
https://www.inf.ethz.ch/personal/fukudak/cdd_home/index.html still points to 0.94i. |
comment:28
It has the author's approval. However, the author is considering to abandon it if the recent acquisiton of github turns the platform the wrong way. |
comment:29
In any case I think we would prefer to have a link to an official tarball rather than an attachment - especially for distros (well me in Gentoo in particular). |
comment:30
Thanks for your work, that greatly simplifies packaging :) Every patch replaced by something upstream supported is a win. Could you maybe ask them to make a note on the cddlib homepage? That would help people that stumble there and also make it seem a bit more legit. |
comment:31
Sorry, I wasn't aware of the details of the SPKG upgrade process. Actually, the checksums are for the tarball on github. |
This comment has been minimized.
This comment has been minimized.
The latest cddlib release |
comment:32
Attachment: cddlib-0.94j.tar.gz Replying to @timokau:
This is now cddlib/cddlib#18. |
comment:33
A link is enough. If we are going to use the tarball https://github.com/cddlib/cddlib/releases/download/0.94j/cddlib-0.94j.tar.gz then spkg-src is not needed. As it been prepared with |
comment:34
Replying to @kiwifb:
The tarball came out of |
comment:35
Replying to @saraedum:
Good enough I guess, although I never used that target. The question about intermediate autotools files is related to |
comment:36
update milestone 8.3 -> 8.4 |
Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:
|
comment:38
trivial merge |
comment:39
Don't modify positively reviewed tickets unless you have a reason. There was no merge conflict, so no need to do anything. I'll ignore the superfluous merge. |
New commits:
|
New commits:
|
comment:43
The trac plugin automatically updates the commit field to the branch head, even though this is not the version that was merged (=c4b0f10c614c7bfa5efaa8ab3908ccb687e992e4). |
comment:44
I needed a clean patch against 8.3 for Debian and Conda distribution so I decided to push the merge so that I can say in the packaging scripts that this is actually the patch exactly as it is going into upstream. I am sorry if this caused some trouble for you. I didn't see the harm of merging in develop. What is the problem other than some noise here on the ticket? |
comment:45
The problem is that I find that the branch has changed after running the buildbot, when I'm about to close the ticket.... |
comment:46
I see. Sorry for that then. |
and use cddexec instead of cdd_both_reps. cddexec should be faster and is, according to upstream, "done the right way". The change to cddexec and the version upgrade make some numerical issues visible, so some things that worked before are not supported anymore. Additionally, we now require an inexact redundancy-free representation to be convertible back and forth between V and H representation (without loss of vertices/inequalities) before we trust it. Some standard examples, fail this check. I can not really make the call on whether these are bugs in cddlib or numerical issues that could not possibly be fixed (as I do not understand this well enough.)
So why this change? Mostly because we do not patch cddlib anymore and just use it as is. It is also faster in some cases (that's what upstream cared most about when the cddlib author heard about our cdd_both_reps,) e.g.,
went from 993ms to 246ms
Some other examples see minor slowdowns because of the additional transformation that we compute initially to make sure that the conversion back and forth worked out.
Note that we also don't need the patched random number generator as upstream now uses a portable rng.
Tarball: https://github.com/cddlib/cddlib/releases/download/0.94j/cddlib-0.94j.tar.gz
CC: @kiwifb @vbraun @timokau
Component: packages: standard
Author: Julian Rüth
Branch/Commit: u/saraedum/25344 @
9b7ee99
Reviewer: Erik Bray
Issue created by migration from https://trac.sagemath.org/ticket/25344
The text was updated successfully, but these errors were encountered: