Bug : bzr pull doesn't display error on inaccessible archives

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Erik de Castro Lopo wrote:
> HI all,
> 
> I've just updated to the latest version from :
> 
>      http://bazaar-ng.org/bzr/bzr.dev/
> 
> However if I now use this new version to do a "bzr pull" on an archive
> which seems to have disappeared off the net it get:
> 
>     > bzr pull
>     Using saved location: http://mysite.verizon.net/sward.dev/projects/bzr_difftools/
>     0 revision(s) pulled.
> 
> whereas I used to get:
> 
>     Using saved location: http://mysite.verizon.net/sward.dev/projects/bzr_difftools/
>     bzr: ERROR: http error 502 probing for http://mysite.verizon.net/sward.dev/projects/
>     bzr_difftools/.bzr/branch-format
> 
> If a archive is inaccessible, it shouldn't report that it sucessfully pulled
> 0 branches.
> 
> Erik

When I try to access that particular site, I now get:

bzr: ERROR: pycurl.error: (52, 'Empty reply from server')

I also get a big nasty traceback, because we don't have explicit
handling of the 'Empty reply from server' return code.
Eventually the server must have woke up, because I started getting
'These branches have diverged'.

At one point we were supressing random pycurl exceptions, but with my
latest http changes we are raising them again.

502 is a "Bad Gateway" response:
http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html

So it seems like it might be something misconfigured on your machine.

I agree that it shouldn't report it successfully pulled nothing. I
believe this is now fixed. Can you try it? I realize having branches
disappear is not an every day occurrence, so you may have to wait for it
to happen again.

I know I get an almost reasonable error trying to pull from a
non-existant host.
% bzr pull http://not-a-real-host.com/bzr/bzr.dev
bzr: ERROR: Connection error: curl connection error (Success) on
http://not-a-real-host.com/bzr/bzr.dev/.bzr/branch-format

The '(Success)' portion is probably confusing (though curl is telling us
os.errno == 0.)

Anyway, there was a bug in that we were squelching pycurl exceptions,
though I don't know how that would have translated into '0 revisions
pulled'.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFExmXVJdeBCYSNAAMRAh30AJoDew+YI/qGkU0rkVKE3XAdbjhrswCgi5qt
ai650ZEU47WIxUzIgDR5t34=
=L2hL
-----END PGP SIGNATURE-----