Collecting cvsps patches

Since there are has been some work done here and there on cvsps, but upstream does not seem to have time to issue a new release, I have started to collect the patches I found.

I guess this is a good place for a heads-up: if you know of any other bugfixes or feature patches to cvsps, I'd like to hear about it, so I can add it to my repo.

Not that the master branch is an octopus merge of all works in there, including my preliminary work on multiple-tag support, so for now you may want to do your own mix.

For now it has:

* bugfixes and such:

Anand Kumria:
      FreeBSD isn't evil - just misguided

Linus Torvalds:
      Increase log-length limit to 64kB
      Improve handling of file collisions in the same patchset
      Fix branch ancestor calculation

Yann Dirson:
      Cleanup the tag handling to simplify multi-tag handling
      Dependency handling

* features

Yann Dirson:
      Allow to have multiple tags on a single patchset.

Hi, Yann!

On Sun, 2006-06-11 at 14:27 +0200, Yann Dirson wrote: 

That's great news. Thank you! I'm sending four patches - two compile fixes for recent regressions and two patches fixing DNS resolution on 64-bit systems - one on Linux and the other on other OSes.

Hi Pavel,

On Tue, Jun 13, 2006 at 12:35:10AM -0400, Pavel Roskin wrote:

Thanks, applied the two latter fixes, and folded the compile fix into the patch it should have been part of :)

For the dependency stuff, I added the -Y flag to hte makedepend invocation. This produces many warnings, but at least does the job right.

I'll push the whole once I have finished a little work on another branch.

Please excuse any delays due to the recent birth of our 3rd son ;)

Best regards,

Yann Dirson wrote:

Congratulations. I'm really glad someone has taken some time to collect these patches. I feel really bad about 'abandoning' (for now only, I hope) the project.

At the very least, I should put some mention in the web page, is there some text you want me to put up there, and/or a like?

David

Hi David,

On Tue, Jun 13, 2006 at 09:46:46AM -0400, David Mansfield wrote:

I have setup a Q&D page at http://ydirson.free.fr/en/software/scm/cvsps.html to link to.

I'll expand it later with more information.

Best regards,

On Tue, 2006-06-13 at 17:15 +0200, Yann Dirson wrote:

Thank you!  By the way, CVSps is capitalized as CVSps, not CvsPs.

It would be great to install the gitweb interface so that the patches could be viewed online.

On Tue, 13 Jun 2006, Yann Dirson wrote:

Will you be able to set up gitweb on that site, perhaps? 

Also, I guess that patch I posted in the "cvsps wierdness" thread (yeah, yeah, bad spelling, but it wasn't me who started the thread) should probably be added, since it fixed at least one test-case.

Although it should probably get more testing with a bigger and more real-life repository..

(David: in case you didn't follow the git list, it makes "compare_patch_sets_by_time()" compare by members _first_, and by time _second_, so that if a file revision shows that a patchset was before another one, it will use that as the primary sort order).

			Linus

On Tue, Jun 13, 2006 at 08:47:15AM -0700, Linus Torvalds wrote:

No, CGI's on this hosting platform are restricted to ISP-provided ones.

Do you think it would be adequate to have this repo hosted on kernel.org ?

Added, in the to-check branch.  At least the idea is sound.

Anyway, there are quite a number of hairy repos out there, which appear to cause problem to cvsps. I'll see if I can identify more problems precisely - I have already collected a handful of them to http://bugs.debian.org/cvsps.

Best regards,

Yann Dirson wrote:

I have added above link to GitWiki, at InterfacesFrontendsAndTools, and at GitLinks.

I have now added all patches gathered from here, as well as those in the current Debian package. Still have to dig other distros.

Note the log-length limit increment is being superceded by a dynamic allocation patch.

Repo URL:  http://ydirson.free.fr/soft/git/cvsps.git

==============================

	* master has:

Alexander Litvinov:
      Handle cvs repo with modules

Anand Kumria:
      FreeBSD isn't evil - just misguided

David D. Kilzer:
      cvsps: should ignore TRUNK branch if it exists in log
      Dynamically allocate the log buffer to prevent warning messages

Linus Torvalds:
      Increase log-length limit to 64kB
      Improve handling of file collisions in the same patchset
      Fix branch ancestor calculation

Pavel Roskin:
      Use __linux__ conditional, not LINUX.
      Use INADDR_NONE instead of -1 to check inet_addr() result

Roberto C. Sanchez:
      Diff opts typo fix.

Yann Dirson:
      Call cvs with -q flag when fetching the log
      Dependency handling

	* to-check additionally has:

Linus Torvalds:
      Make time ordering less important than revision ordering

	* dev branches still have incomplete stuff, so I don't list it
	here.

On Mon, Jun 12, 2006 at 10:16:55AM +1200, Martin Langhoff wrote:

Rather, I have carefully hidden it in a previous post with an unrelated Subject line:)

http://ydirson.free.fr/soft/git/cvsps.git

Best regards,

On Mon, 12 Jun 2006 00:42:05 +0200, Yann Dirson wrote:

I think you need to chmod +x hooks/post-update

and then run 'git-update-server-info'.

Cheers, Anand

On Mon, Jun 12, 2006 at 11:27:37AM +0000, Anand Kumria wrote:

Unfortunately, I only have FTP access to push to this site, so I have to run git-update-server-info myself, and occasionally forget. I'll have to bring up-to-date my old cg-ftppush script some day :)

Best regards,

Yann Dirson wrote:

Or extend git to allow push/pull also via ftp?

Jakub Narebski <jnareb@gmail.com> writes:

Patches welcome.

Yann Dirson <ydirson@altern.org> writes:

When I only had "serving HTTP, published via FTP" ISP accounts, I essentially ended up having a bare repository on my local machine, update it with real git-push (with the post-update hook there), and copied that to the ISP via lftp batch.

Your master branch is not compilable. This is the fix for it.

---
 cache.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

On Thu, 2006-06-15 at 09:21 +0700, Alexander Litvinov wrote:

Actually, revision 5ecebc5064df0fd578dbf4b5ba5255e9af8cda7a renamed "tagnames" to "link" in cvsps_types.h, so I think the same should be done in cache.c.

Also, you can check the definition of list_entry, you'll see that the member (argument 3) is cast to the pointer to the requested type (argument 2). I don't think it's safe to cast name to a pointer to TagName.

That said, having non-compilable repository defeats the purpose of collecting the patches to cvsps. Shouldn't only compilable patches be allowed in the repository?

You are right. My patch is wrong. I have tried to import real cvs repo and cvsps broke with sigsegv.

On Wed, Jun 14, 2006 at 11:08:32PM -0400, Pavel Roskin wrote:

Right, it is entirely my fault, I guess I got distracted at that time.

I have change the layout of the repo to avoid such mishappenings in the future. The master branch holds the fixes, and new features or intrusive fixes go into their own branches (especially since none of them are complete yet).

Fixes that may need further thought are in the to-check branch.

Best regards,

On Thu, Jun 15, 2006 at 09:21:34AM +0700, Alexander Litvinov wrote:

Right, it was fixed already.

Best regards,