[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#603463: apache2-mpm-worker: Seems to send incorrect multi-page reply



Op 16-11-10 23:50, Stefan Fritsch schreef:
> On Sunday 14 November 2010, Bas Wijnen wrote:
>> Everything normally works fine. However, in some cases (I don't
>> know when exactly), I get an invalid reply from my local archive.
>> I've captured the traffic with wireshark and attached it.
> 
> Do you have mod_cache enabled in apache? If yes, this is probably 
> https://issues.apache.org/bugzilla/show_bug.cgi?id=45341
> and you could try to disable mod_cache for that virtual host.

No, I don't have it enabled.

>> The problem seems to be in packet 19. It contains 3 HTTP parts, the
>> second of which does not seem to have data in it (according the
>> wireshark's parser). However, when looking at that part, it not
>> only contains data: it also contains the next HTTP header.
> 
> The problem is that there is a 304 response which should not happen 
> for an unconditional request. But the 304 response not having any body 
> is OK (it doesn't have a content-length header either).

Yes, my problem with the packet isn't that it doesn't have data, but
that (according to wireshark's parser) the data that really is there is
not parsed as such, plus the header for the next packet is not parsed as
a header.

The "missed" (by the parser) header seems to happen to apt as well,
because it starts waiting until it times out for a reply that never
comes. In total, there are 10 requests, and if I look at the data there
are 10 replies, but one of them is not parsed as such, so it waits for
the last one.

I'm not familiar enough with http traffic to see why that header is missed.

> This traffic 
> goes to port 80, so I think the problem has nothing to do with apt-
> cacher-ng.

Indeed, that was not a very clever suggestion. The reason I thought of
this (apart from being confused about which data stream this was) is
that I tried to simplify the problem by removing all other repositories
from my sources.list. That resulted in the bug disappearing. This is why
I'm not so sure if this is (only) a bug in apache, because (the locally
running) apache shouldn't notice at all if apt also connects to some
other repository on a different port on a different host.

Thanks,
Bas

Attachment: signature.asc
Description: OpenPGP digital signature


Reply to: