Darcs fails if big record is commited - around 2100 new files which totals to 20
MB of data.
Faulty repo can be downloaded at: http://zuzia.hopto.org/~radekg/faultyrepo.zip

darcs --exact-version gives:
darcs compiled on Nov 14 2005, at 22:48:32
# configured Mon Nov 14 22:35:33 USMST 2005
./configure --disable-mmap

Context:

[update web page to reflect 1.0.4 as latest stable source.
David Roundy <droundy@darcs.net>**20051113210144]

[TAG 1.0.4
David Roundy <droundy@darcs.net>**20051113134431]

Forgot to add: Darcs fails with unexpected error - without any message. Standard
windows dialog "Send details" pops up. Darcs leaves lock file behind.

Adding to the bug report that Will has also seen this.

Will Glozer wrote:
> I've been experiencing the same problem importing a very large tree
> with many binary files as a single changeset.  I've been testing with
> the following archive contents:
>
>  http://prdownloads.sourceforge.net/jboss/jboss-4.0.3SP1.tar.bz2?download

I've tested this tarball (which seemed faster to download) on linux and
haven't had a problem, and I think Will said it worked fine on MacOS X.

I ran

darcs initialize
darcs record -al -m foo

Would that be enough to trigger the bug on Windows?

It would also be a good idea to just try running on linux or MacOS mmap
disabled, just to increase the similarity with the windows scenario.  (But
I don't have time for this today--I wasted some time fixing a bug in the
nosy keywords, which should have added me to the nosy list for this bug.)
-- 
David Roundy
http://www.darcs.net

2005/11/16, David Roundy <bugs@darcs.net>:
>
> David Roundy <droundy@darcs.net> added the comment:
>
> Adding to the bug report that Will has also seen this.
>
> Will Glozer wrote:
> > I've been experiencing the same problem importing a very large tree
> > with many binary files as a single changeset.  I've been testing with
> > the following archive contents:
> >
> >  http://prdownloads.sourceforge.net/jboss/jboss-4.0.3SP1.tar.bz2?download
>
> I've tested this tarball (which seemed faster to download) on linux and
> haven't had a problem, and I think Will said it worked fine on MacOS X.
>
> I ran
>
> darcs initialize
> darcs record -al -m foo

Tested on my repo - yes.

Radek.

This fails with 1.0.3 as well.  In fact, with 1.0.3 it says:

 Kanga:~/t$ darcs record -al -m foo
Darcs needs to know what name (conventionally an email address) to use as the
patch author, e.g. 'Fred Bloggs <fred@bloggs.invalid>'.  If you provide one
now it will be stored in the file '_darcs/prefs/author' and used as a default
in the future.  To change your preferred author address, simply delete or edit
this file.

What is your email address? foo
darcs.exe: internal error: stg_ap_v_ret
    Please report this as a bug to glasgow-haskell-bugs@haskell.org,
    or http://www.sourceforge.net/projects/ghc/

That is the same error message that I've often gotten with this bug:

http://otherbugs.darcs.net//Ticket/Display.html?id=339

It might be related to this bug:

http://otherbugs.darcs.net//Ticket/Display.html?id=556

It is definitely the same as these two bugs:

http://sourceforge.net/tracker/index.php?func=detail&aid=1194393&group_id=8032&atid=108032
http://sourceforge.net/tracker/index.php?func=detail&aid=1194392&group_id=8032&atid=108032

Hopefully radekg's contribution of a reproducible test case will spur progress
on this bug.

However, since this is not a regression from 1.0.3 -> 1.0.4, and is instead a
bug that has been in place for time immemorial, it should not hold up the darcs
1.0.4.1 release.

By the way, the workaround is "have smaller patches".

Thanks Zooko!
I'll have 1.0.5rc1 ready on friday.

Hooray!  Fixed:

http://sourceforge.net/tracker/index.php?func=detail&aid=1194393&group_id=8032&atid=108032

Now to fix darcs on w32 we have to recompile ghc and then use ghc to recompile
darcs for w32.  Then I'll package it up.

Yay!

Thank you for getting this fixed zooko!! I can confirm that the jboss import
test case works with GHC's ghc-6-4-stable branch.  I've sent in a few patches
that are neccessary to compile with the latest GHC, and it may be a few days
before the daily builds are working.  In the meantime I've made a temporary
1.0.5rc1 binary that contains the fixes:

http://glozer.net/darcs/darcs-1.0.5rc1-win32.zip

Happy to help, but for the record it was radekg that provided a reproducible 
test case and simonmar that fixed it.

I'll package your latest binary (which means copying it over darcsdir-cygwin/
realdarcs.exe and copying it over darcsdir-win32/darcs.exe and then zipping, 
tarring, and uploading the results).

Shouldn't we bump the version number to darcs-1.0.5rc2?

On Sun, Dec 04, 2005 at 01:21:33PM +0000, Zooko wrote:
> Shouldn't we bump the version number to darcs-1.0.5rc2?

In a minute... :-)

I'll apply some more patches and tag 1.0.5rc2 later today,
and then there will be a 1.0.5 in just a day or two, when a
few more people have had the chance to build and run the test.

2005/12/4, Zooko <bugs@darcs.net>:
>
> Zooko <zooko@zooko.com> added the comment:
>
> Happy to help, but for the record it was radekg that provided a reproducible
> test case and simonmar that fixed it.

You are to kind. You've made a great deal of work. Thanks.

> Shouldn't we bump the version number to darcs-1.0.5rc2?

Shouldn't this go to stable? Or will it?

Cheers,
  Radek.

Hm.  We got a similar error message just now on a buildbot:

http://buildbot.darcs.net/builders/zooko%20allmydata%20slave3%20Linux-2.6.15%20Ubuntu-6.06.2%20LTS%20i386/builds/47/steps/test/logs/stdio

Resetting to resolved from msg150.

Also: Hi Zooko,

I believe you were referring to this error message:

darcs.exe: internal error: stg_ap_v_ret
    Please report this as a bug to glasgow-haskell-bugs@haskell.org,
    or http://www.sourceforge.net/projects/ghc/

Normally, I'd split this off into a new bug report (I like to treat rotting
zombies as new people), but I notice that this was for a 2006 version of Ubuntu,
which means old GHC to me so I'm going to assume that this was fixed in one of
the later GHCs since your last report.  If you run into this again, please open
a new ticket.

Thanks!