[cl-debian] yet another versioning change?
Hello, For the last week I've been trying to upgrade slime to the newest version and failed. slime-upstream works fine, but I cannot pull the changes into slime. I've send emails to the darcs-users list and debian-devel, but no-one offered any clues as to how I can solve this. I fear that basically darcs is not suited to debian package management because we often introduce 'conflicts' and 'doubleganger patches' (upstream includes a patch we've send). darcs seems to just break down under these conditions. I've been redoing my work on sbcl for the last year with bzr and so far it did not blow up ;-S. Using tailor to convert from darcs to bzr does seem to take quite some time and I have not been able to check if you can do a merge between a <foo>-upstream and a <foo> branch after converting them with tailor. I fear not and that would mean we would have to restart all repositories. again. Opinions? Groetjes, Peter -- signature -at- pvaneynd.mailworks.org http://www.livejournal.com/users/pvaneynd/ "God, root, what is difference?" Pitr | "God is more forgiving." Dave Aronson|
Hello! Sorry for the long mail, it contains a bit of output to explain my problem. On Fri, 04 Aug 2006 23:01:11 +0200, Peter Van Eynde wrote:
For the last week I've been trying to upgrade slime to the newest version and failed. slime-upstream works fine, but I cannot pull the changes into slime.
If you remember, I have a similar problem with my arnesi repositories: every time I pull from arnesi-upstream into arnesi, the pull takes at least 90min (with the darcs process going up to 99% of the CPU). This was probably due to some `darcs unrecord` I performed on arnesi (not -upstream) and it's reproducible with a fresh copy: ===== luca@gismo:~$ cd test/ luca@gismo:~/test$ ls arnesi* ls: arnesi*: No such file or directory luca@gismo:~/test$ time darcs get \ http://cl-debian.alioth.debian.org/repository/lcapello/arnesi-upstream Copying patch 285 of 285... done! Applying patch 285 of 285... done. Finished getting. real 0m18.841s user 0m0.948s sys 0m0.040s luca@gismo:~/test$ time darcs get \ http://cl-debian.alioth.debian.org/repository/lcapello/arnesi Copying patch 307 of 307... done! Applying patch 307 of 307... done. Finished getting. real 1m44.539s user 0m1.032s sys 0m0.112s luca@gismo:~/test$ cd arnesi-upstream/ luca@gismo:~/test/arnesi-upstream$ time darcs pull --all \ http://common-lisp.net/project/bese/repos/arnesi_dev Pulling from "http://common-lisp.net/project/bese/repos/arnesi_dev"... Finished pulling and applying. real 0m1.930s user 0m0.220s sys 0m0.004s luca@gismo:~/test/arnesi-upstream$ cd ../arnesi luca@gismo:~/test/arnesi$ time darcs pull ../arnesi-upstream/ Pulling from "../arnesi-upstream"... withSignalsHandled: Interrupted! real 11m48.508s user 11m20.451s sys 0m1.624s luca@gismo:~/test/arnesi$ =====
I've send emails to the darcs-users list and debian-devel, but no-one offered any clues as to how I can solve this.
I read them. I asked once on #darcs@irc.freenode.net about my problem with arnesi, but no one could solve it :-(
I fear that basically darcs is not suited to debian package management because we often introduce 'conflicts' and 'doubleganger patches' (upstream includes a patch we've send). darcs seems to just break down under these conditions.
Well, actually I think for me darcs is OK because I work differently AFAIU. I don't record the debian specific patches as darcs patches, but I apply them through dpatch instead (and `darcs record` the dpatch files). In this case, if a patch is applied upstream, it's just a question of removing the corresponding dpatch file. I know that this isn't probably the aim of an SCM, but it works quite well for me, so I'm not inclined to change it.
I've been redoing my work on sbcl for the last year with bzr and so far it did not blow up ;-S. Using tailor to convert from darcs to bzr does seem to take quite some time and I have not been able to check if you can do a merge between a <foo>-upstream and a <foo> branch after converting them with tailor. I fear not and that would mean we would have to restart all repositories. again.
IMHO this won't be a problem as far as the debian/changelog documents all the changes.
Opinions?
I'm for one SCM for all the CL-Debian packages, but we need to deal with different upstream SCMs and in my case most of them uses darcs. I don't really know what the Right Thing™ is :-( Thx, bye, Gismo / Luca
Alle Friday 04 August 2006 23:01, Peter Van Eynde ha scritto:
I've been redoing my work on sbcl for the last year with bzr and so far it did not blow up ;-S
I've uploaded my test bzr repositories to alioth. You can do a branch with To get the repository: $ bzr branch http://cl-debian.alioth.debian.org/repository/pvaneynd/bzr-moved/sbcl/ To add files: $ bzr add To commit changes: $ bzr commit -m 'added unicode support' To see what changed: $ bzr diff To unify with another tree: $ bzr merge <location> after which you have to do a commit to commit the merge. (Which is a bit nicer then in darcs) Sending your changes via email seems to be coming soon: http://bazaar-vcs.org/SubmitByMail I have also make a yet-untested http://cl-debian.alioth.debian.org/repository/pvaneynd/bzr-moved/bzr-build.s... script. It uses a special commit message of UPSTREAM_sbcl_0.9.15.0 to act as 'tag'. I expect bzr to gain tags soonish (http://bazaar-vcs.org/BzrTagging), but with this trick I can handle my work-flow as is. The repositories that darcs can handle I'll just leave as-is for the moment. Groetjes, Peter -- signature -at- pvaneynd.mailworks.org http://www.livejournal.com/users/pvaneynd/ "God, root, what is difference?" Pitr | "God is more forgiving." Dave Aronson|
Il giorno mar, 08/08/2006 alle 11.30 +0200, Peter Van Eynde ha scritto:
Sending your changes via email seems to be coming soon: http://bazaar-vcs.org/SubmitByMail
[...]
I have also make a yet-untested http://cl-debian.alioth.debian.org/repository/pvaneynd/bzr-moved/bzr-build.s... script. It uses a special commit message of UPSTREAM_sbcl_0.9.15.0 to act as 'tag'. I expect bzr to gain tags soonish (http://bazaar-vcs.org/BzrTagging), but with this trick I can handle my work-flow as is.
If you're going to do some tests, I'd suggest to try Mercurial: http://selenic.com/mercurial I've used it for some time (with small source trees), without problems and with a lot of pleasure. Some goods: * it is fully supported by Tailor; * it supports tagging; * it supports sending and importing changeset via email; * it does scale. About scaling: Mercurial was created as an alterative to Git for the development of Linux, and their speed is almost the same (but Mercurial is written in Python instead of C, and requires a lot less bandwidth and disk space for transmitting and managing repositories). It is currently hosting the Linux source tree (see http://kernel.org/hg/linux-2.6/ ) and has been selected as the official distributes SCM system for OpenSolaris. Oh, it also offers a nice webserver/CGI for browsing the repositories, generating RSS feeds for changelogs, and eventually push'ing changesets through HTTP/HTTPS. Regards, alceste -- Alceste Scalas <alceste@muvara.org>
On Tue, Aug 08, 2006 at 10:44:24PM +0200, Alceste Scalas wrote:
If you're going to do some tests, I'd suggest to try Mercurial.
See also http://en.wikipedia.org/wiki/Comparison_of_revision_control_software -- Trent Buck, Student Errant
participants (4)
-
Alceste Scalas -
Luca Capello -
Peter Van Eynde -
Trent Buck