RE: lbwww crashing after the deletion of an idle timed out HThost

(Sorry I previously answer on the wrong message)

I got the same problem and I have comment this in HTHost.

But back to Sven Laaks whet belongs to Sven Laaks!

see "Subject: Robot. Crash or memory problems after 12 hours"

When I read it's first mail I found it funny.

Few after the programs stop after exactly 12h.

I've writed to the list from work but my mail never been in the list.

I'can't anymore write in the list from work. Only from home.

Don't know why.

Sven was in copy. Don't know if he ever got my answer.

Here it is:

Hi Sven , hi all.

Sven Laaks wrote:

> I modified the example robot, so that it checks out a webpage 

> in given intervalls (e.g. every 10 minutes). 

> My first problem is, that the robot terminats after exactly 12h 

> with "segmentation fault (core dump)". 

> I work on a program that does the same thing but which havn't 

> been designed from the robot example.

I'm now tracking this 'exactly 12h' issue. 

[...]

The 12h is in the lib. It is in seconds unit. 

Search for 4320 in the source files and you will find what happens.

I gonna first try to disable this feature for I miss time.

Maybe add a 'HTHost_clearChannel()' when the host timeout expires 

could help.

> On other pages it runs 24h and more. 

It depends on whether the host is idle or not.

With multiple requests on an HTTP 1.1 host you can make it never idle 

and then you don't have this problem.

Michel. 

 

-----Original Message-----
From: Azzurra Pantella [mailto:azzurra.pantella@netikos.com]
Sent: Tuesday, December 18, 2001 4:27 PM
To: www-lib@w3.org
Subject: Fw: lbwww crashing after the deletion of an idle timed out HThost


 
Hi all,
after having  submitted  many GET (using HTLoad_toChunk) requests to the
same HTTP1.1 host, our program crashed and we got some purify memory errors
in libwww code (we were running a .purify exe) .
After a core and a log analysis  we are quite sure that the crash is due to
a libwww problem after the deletion of an idle timed out host.
We noticed that in the HTHost.c module, while searching the HTHost cache
within the function HTHost_new(),
a check is performed in order to delete old (older than  12 h ) and idle
(empty pipeline) HTHost objects. The host was canceled and removed from the
HostTable but apparently there was already an event connected with this host
(a CLOSE event?)   in the "EventOrder" list waiting to be scheduled by
EventOrder_executeAndDelete() (cfr. HTEvtLst.c).
The fields of the HTevent pointed by "pres->event" were previously freed and
a crash resulted.
Has anybody observed similar problems ?
(We attempted a fix if u are interested)
 
Thanks,
            Azzurra
 
 
Azzurra Pantella
Mobile Applications Area
NETikos S.p.A.
Via Matteucci 34/B
56124 PISA
Tel.: + 39 050 968526
Fax: + 39 050 968525
e-mail:  azzurra.pantella@netikos.com <mailto:azzurra.pantella@netikos.com> 
Internet: www.netikos.com <http://www.netikos.com>