Re: OpenSolaris hg clone fails for me

[christos%astron.com@localhost (Christos Zoulas)]

In article <20090324220136.4da8e9f3@michael>,
Michael Smith  <smithm%netapps.com.au@localhost> wrote:
>-=-=-=-=-=-
>
>On Sun, 22 Mar 2009 13:53:03 +0000 (UTC)
>christos%astron.com@localhost (Christos Zoulas) wrote:
>
>> In article <20090322195909.653c88bd@michael>,
>> Michael Smith  <smithm%netapps.com.au@localhost> wrote:
>> >-=-=-=-=-=-
>> >
>> >On Sun, 22 Mar 2009 13:13:14 +0900
>> >Masao Uebayashi <uebayasi%gmail.com@localhost> wrote:
>> >
>> >> I get following error on NetBSD/amd64 5.99.5 + Mercurial 1.1.2. 
>I'm curious
>> >> if this works for your environment...
>> >> 
>> >> % hg clone ssh://anon%hg.opensolaris.org@localhost/hg/onnv/onnv-gate
>> >> destination directory: onnv-gate
>> >> requesting all changes
>> >> adding changesets
>> >> adding manifests
>> >> transaction abort!
>> >> rollback completed
>> >> abort: premature EOF reading chunk (got 607178 bytes, expected 2766212)
>> >> remote: Read from remote host hg.opensolaris.org: Connection reset by peer
>> >
>> >Works for me with:
>> >
>> >NetBSD 4.0_RC3 hg  0.9.3
>> >
>> >It looked to be working on another system (NetBSD 4.0.1/hg 1.0.2) but I
>> >stopped that because hg.opensolaris.org was sending data slowly and I
>> >didn't want to cause more problems by doing two downloads as once.
>> >
>> >But it is slow going. I wonder if the line into hg.opensolaris.org is
>> >up to the task.
>> 
>> There are bugs in the Solaris hg side with 64 bit timestamps. I managed
>> to corrupt it, committing from the latest NetBSD+Mercurial. Fetching though
>> has always worked for me.
>
>On my NetBSD 4.0_RC3 machine I upgraded to hg 1.2.1
>I am using Python 2.5.2 from pkgsrc.
>
>I cloned a new repository from hg.opensolaris.org. I added a new file
>and comitted it. I didn't see any problems. Can you tell me what steps
>you followed to corrupt the repository? I would like to file a bug with
>the mercurial project for this if it can be shown to be reproducable.

I use current where time_t is 64 bits. Old versions of mercurial apparently
don't like this.

christos