https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
Bernard Spil changed:
What|Removed |Added
Status|NEW |RESOLVED
Resolution|---
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #14 from Yann Ylavic ---
Follow up in r1840002, simpler to test latest trunk anyway.
--
You are receiving this mail because:
You are the assignee for the bug.
-
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
Yann Ylavic changed:
What|Removed |Added
Status|NEEDINFO|NEW
--- Comment #13 from Yann Ylavic
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #12 from Yann Ylavic ---
Nevermind, I think I can reproduce with large download and some rate limit on
the client side. Let me see what happens...
--
You are receiving this mail because:
You are the assignee for the bug.
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #11 from Yann Ylavic ---
It would be interesting to know which filters are pending in
ap_filter_output_pending() too (i.e. print f->frec->name in the loop).
--
You are receiving this mail because:
You are the assignee for the
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #10 from Yann Ylavic ---
I don't see how ap_filter_output_pending() can loop indefinitely..
Can you please step (or next) there from gdb, with "set scheduler-locking on"
to stay in the same thread (and preferably symbols if
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #9 from Bernard Spil ---
Hi Yann,
-fno-strict-aliasing is enabled by default. Build log of an earlier version
here:
https://brnrd.eu/poudriere/data/112libre-default/2018-08-05_22h41m54s/logs/apache25-devel-2.5.1.20180529.log
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #8 from Yann Ylavic ---
OK thanks, if you don't already, can you try to compile with
-fno-strict-aliasing?
Aliasing is known to break with APR_RING...
--
You are receiving this mail because:
You are the assignee for the bug.
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #7 from Bernard Spil ---
Created attachment 36129
--> https://bz.apache.org/bugzilla/attachment.cgi?id=36129=edit
stepping through util_filter.c 1012-1032
Stepping through I see that I'm switching between the threads that are
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #6 from Bernard Spil ---
(In reply to Yann Ylavic from comment #4)
> Bernard, is this with high concurrency scenario or it can be reproduced with
> a single/few requests?
>
> Looking at the stacktrace, the most probable looping
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #5 from Bernard Spil ---
(In reply to Eric Covener from comment #3)
> (In reply to Bernard Spil from comment #0)
> > Created attachment 36126 [details]
> > gdb thread apply all backtrace
> >
> > Somewhere between 2018-06-20 and
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #4 from Yann Ylavic ---
Bernard, is this with high concurrency scenario or it can be reproduced with a
single/few requests?
Looking at the stacktrace, the most probable looping threads are numbers 60 to
62, thus
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
Eric Covener changed:
What|Removed |Added
Status|NEW |NEEDINFO
--- Comment #3 from Eric
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #2 from Bernard Spil ---
The version I roll back to to make this work (build date is also trunk date):
> # httpd -V
> Server built: Jun 20 2018 18:28:30
> Server's Module Magic Number: 20180606:1
First version I have exhibiting
https://bz.apache.org/bugzilla/show_bug.cgi?id=62668
--- Comment #1 from Bernard Spil ---
The version I roll back to to make this work:
# httpd -V
Server version: Apache/2.5.1-dev (FreeBSD)
Server built: Jun 20 2018 18:28:30
Server's Module Magic Number: 20180606:1
First version I have
15 matches
Mail list logo