Google + Mercurial is seriously broken!

This has been a problem from the beginning of this project, but over the past two days the problem has become chronic.
martin@martin-Inspiron-620:~$ cd prop martin@martin-Inspiron-620:~/prop$ hg clone https://code.google.com/p/propgcc/ destination directory: propgcc requesting all changes adding changesets adding manifests transaction abort! rollback completed abort: connection ended unexpectedly martin@martin-Inspiron-620:~/prop$ hg clone http://code.google.com/p/propgcc/ destination directory: propgcc requesting all changes adding changesets adding manifests transaction abort! rollback completed abort: connection ended unexpectedly martin@martin-Inspiron-620:~/prop$
Comments
I've had to try up to 5 times. I feel the pain.
The number of files is probably responsible for most of the trouble.
I'll try cloning it and changing to git to see if it works any better.
I gave up at around 20 to 30 tries in rapid succcession.
I use SVN in a multi-developer environment and the rule of thumb is "update to head" regularly. Eclipse manages SVN integration very well, eliminating much of the oddball nature.
The big difference between SVN and HG&GIT is that you needn't clone the entire history of civilisation to get the latest rev. It's a thin client model, whereas HG&GIT aren't, and that compounds the problem you are experiencing.
The PropGCC repo is several GB in size, and needlessly so.
It HG is failing though. Then there's reason to swtich.
Thanks,
I do find branching to be problematic and prefer working from the trunk, to force the responsibility of merging onto the submitter, rather than consumer.
I feel it is the submitter's responsibility to ensure their changes don't break the build, and forcing them to merge on checkin is one way to prevent that.
Otherwise, if someone checks in a conflicting change, I'm forced to integrate their work and I may break their code as a result.
My first experience with checkin on HG resulted in needless hoops WRT merges. I could see the solution plain as day, but it made the process so frustrating.
Maybe it influences some timeouts too?
When giving --debug cloning fails far less often.
Without --debug I never was able to clone propgcc.
Updating the repo (the local clone) without --debug always was no problem here.