Trying to get bzr.dev: Invalid http response

Vincent Ladeuil v.ladeuil+lp at free.fr
Thu Nov 29 06:59:05 GMT 2007


>>>>> "martin" == Martin Pool <mbp at sourcefrog.net> writes:

    martin> On Nov 29, 2007 3:50 AM, John Arbash Meinel <john at arbash-meinel.com> wrote:
    >> -----BEGIN PGP SIGNED MESSAGE-----
    >> Hash: SHA1
    >> 
    >> Vincent Vertigo wrote:
    >> > And by the way, I am using bzr0.92 to do this.
    >> > (I was using 0.91 until an error about knitpack format "forced" me to
    >> > retry with bzr0.92)

Look at your .bzr.log file and please post appropriate parts,
that should help us understand what happens.

    >> >
    >> 
    >> Sounds a bit fishy. Robert did upgrade the branch to packs, which means you
    >> need at least 0.92 to read it.
    >> 
    >> There are a few possibilities:
    >> 
    >> 1) Make sure you can get to:
    >> http://bazaar-vcs.org/bzr/bzr.dev/.bzr/branch/format
    >> 
    >> However, if you are getting far enough to get to *.pack, then bzr should have
    >> been able to get to it.
    >> 
    >> 2) I'm guessing it is complaining because of a partial range request on that
    >> file. We fixed a small bug about this in bzr.dev (so that we issue fewer ranges).
    >> 
    >> I'm very surprised to see it be 503: Service Unavailable.
    >> But it may be that a proxy is falling over because of a complicated range
    >> request, and giving 503 rather than a better error code.

    martin> I had a look at the logs on that server, and in the
    martin> last 4 days apache did not log any 503 errors.  So it
    martin> seems most likely that the error was returned by an
    martin> intermediate proxy server, possibly because our range
    martin> request was too complicated, or possibly because of
    martin> some error on that machine not related to the
    martin> request.  It's conceivable there was an error on
    martin> bazaar-vcs.org severe enough to not be logged, I
    martin> suppose.

    martin> This kind of error would probably ideally be handled
    martin> by printing a warning, waiting a bit, and retrying...

Except for the 'waiting a bit' that's what we do (by the time we
get the error, one can even argue that in fact we have already
wait), I agree about printing (as opposed to write it to the log
as we do now) the warning and we should also mention the file
involved.

        Vincent



More information about the bazaar mailing list