Re: [PATCH] soup-message-io: protect unpause_source by mutex
- From: "Andrew W. Nosenko" <andrew w nosenko gmail com>
- To: Dan Winship <danw gnome org>
- Cc: libsoup-list gnome org
- Subject: Re: [PATCH] soup-message-io: protect unpause_source by mutex
- Date: Thu, 21 Jan 2010 18:11:20 +0200
On Thu, Jan 21, 2010 at 16:29, Andrew W. Nosenko
<andrew w nosenko gmail com> wrote:
> On Thu, Jan 21, 2010 at 15:52, Dan Winship <danw gnome org> wrote:
>> On 01/21/2010 03:37 AM, Andrew W. Nosenko wrote:
>>> On Wed, Jan 13, 2010 at 13:45, Andrew W. Nosenko
>>> <andrew w nosenko gmail com> wrote:
>>>> soup-message-io: protect unpause_source by mutex
>>>>
>>>> Otherwise the the multithreaded program, which pauses message in the
>>>> one (server) thread and unpauses it in the another (worker) thread may
>>>> experience a crash under high workload (and in my case experienced it
>>>> indeed).
>>>
>>> Ping?
>>
>> (So first off, you should file patches in bugzilla, not on the mailing
>> list.)
>
> Hmm... Sorry. I jush have not so good expirience with bugzilla in general...
>
>> Anyway, yes I saw the message.
>
> Thanks. Essencially the "ping" was intended for verify that -- I has
> suspiction that message could be "lost"/"hidden" by some spam filter.
>
>> I haven't look at it in detail yet. Note
>> that the docs state that only SoupSessionSync is thread-safe, and since
>> soup_session_pause_message() only works with SoupSessionAsync,
>
> There are two problems:
>
> 1. It is about SoupServer indeed. Pause/unpause meant by me are
> soup_server_pause_message() / soup_server_unpause_message().
>
> 2. [unrelated to this patch, just for yout info] SoupSessionSync is
> not as thread safe as intended to be, unfortuanlely -- there lives
> some race condition also (and I expirienced crashes there too). But
> in contrast to the SoupServer it was easy worked around by creating
> one SoupSessionSync per one thread, while SoupServer has no such
> ability.
>
>> this
>> probably means you're doing something which is known to not work, and
>> even after fixing this there will probably be other problems too,
>> because none of the async code paths have been tested much for threadsafety.
>
And, again, please take my exuses and apologies -- I didn't have any
intention to insult or bother you in any way. It is perfectly
understandable that you may be busy by something more important for
you (personally or proffecionally -- no mater) or just enjoy holyday
or vacation, and I have no any intention to ruin your plans.
--
Andrew W. Nosenko <andrew w nosenko gmail com>
[
Date Prev][Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]