hc-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Becke <be...@u.washington.edu>
Subject Re: Still problems with HttpRecoverableException and HTTPS
Date Mon, 14 Apr 2003 12:42:19 GMT
In the last few patches this sort of problem has been fixed for 
writing, but not reading.  In particular see:


As others have mentioned, the problem with retrying on the read is that 
it is difficult to tell if the request ever made it to the server.  I 
am inclined to think that if the response could not be read we can 
assume that the request didn't make it.  We should probably get some 
more confirmation on this though, other than gut reaction.

The obvious best solution would be to figure out a way to determine if 
a connection has been closed by the server.  I experimented with this 
for a while but to no avail.  The only idea I have is to try a method 
with no side-effects, like HEAD, on a connection before using it.  We 
wouldn't want to do this for every connection use but it's an idea.

Aurelien, I do not think stepping through debugging will help much.  I 
believe it is a timing issue involving connection reuse and server 
behavior.  Please post some more information on your environment 
(client, server, etc.) and a high level description of the situation 
where this is happening.  I will try to make a reliable test case.


On Monday, April 14, 2003, at 06:16 AM, Aurelien Pernoud wrote:

> I agree with you, although I don't understand why httpclient thinks he 
> wrote
> the request... Most of time the exception is thrown when writing 
> request to
> tomcat, and then retried ok, but not every time.
> I don't really like this "trick" cause how to be SURE the request 
> didn't
> went, in some other case we didn't meet yet ????
> If it can help devs, I've found a way in my app to reproduce the bug 
> every
> time (the request write goes ok but I'm SURE the connection is down), 
> even
> if for this I have to restart tomcat (yes I know, it's weird, but I 
> think
> this all bug is weird).
> The trouble is I can't reproduce it outside of my app (!!!!), and 
> couldn't
> make a testcase that did the same result.
> I went in debug in httpclient, but as I don't know what to look at 
> exactly
> (especially I was looking at the write request that went ok but 
> shouldn't),
> I'm pretty stuck.
> The wire log isn't any help here, it simply says it couldn't read 
> response.
> If anyone has an idea of what to look at while in debug (variables,
> whatever), I'll be able to help ?
> What about creating a bug in bugzilla to test possible patches too, as 
> this
> time It seems I'm not the only one meeting this trouble ?
> Ingo Brunberg a écrit :
>> Your patch looks reasonable to me. It implements what I said: Don't be
>> too smart. The danger, and something, I remember, someone was afraid
>> of is that the request actually did reach the server and gets executed
>> twice. But that's not the case im my scenario and, if I have not
>> misunderstood anything in the previous mails in this thread, noone has
>> ever actually seen such a request reach the server.
>> Regards,
>> Ingo
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: 
> commons-httpclient-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: 
> commons-httpclient-dev-help@jakarta.apache.org

View raw message