MySQL Lists are EOL. Please join:

List:Commits« Previous MessageNext Message »
From:Vladislav Vaintroub Date:October 30 2009 8:15pm
Subject:RE: bzr commit into mysql-5.0 branch (vvaintroub:2829) Bug#31621
Bug#47571
View as plain text  

> -----Original Message-----
> From: Magnus.Blaudd@stripped [mailto:Magnus.Blaudd@stripped]
> Sent: Friday, October 30, 2009 3:56 PM
> To: Vladislav Vaintroub
> Cc: 'Vladislav Vaintroub'; commits@stripped
> Subject: Re: bzr commit into mysql-5.0 branch (vvaintroub:2829)
> Bug#31621 Bug#47571
> 
> Vladislav Vaintroub wrote:
> >
> >> -----Original Message-----
> >> From: Magnus.Blaudd@stripped [mailto:Magnus.Blaudd@stripped]
> >> Sent: Friday, October 30, 2009 10:46 AM
> >> To: Vladislav Vaintroub
> >> Cc: commits@stripped
> >> Subject: Re: bzr commit into mysql-5.0 branch (vvaintroub:2829)
> >> Bug#31621 Bug#47571
> >>
> >> magnus: although I see that vio already have a "net" member, it
> feels
> >> like that is an old hack just to get read and write timeout
> settings.
> >>
> >> Suggest you would instead install a 'vio_pipe_timeout' instead of
> >> 'vio_ignore_timeout' which is currently installed for named pipe
> >> connections. To implement that, you probably would have to add
> >> read/write timeout members to vio, but that is in my opinion better
> >> than
> >> having the whole NET available there. This would make it possible to
> >> change read/write timeout dynamically by calling those functions
> (well
> >> maybe that works already because net->read_timeout is updated in
> >> 'my_net_set_read_timeout').  Well, up to you in the end, but I think
> >> that's the idea how it should work. BTW, "net" would only be
> available
> >> if HAVE_SMEM...
> >
> > Hi Magnus and thanks for detailed review. I see your point and
> generally
> > agree with it. The only thing I'm not comfortable with is having 3
> different
> > methods of handling timeouts :
> > - pass to setsockopt
> > - pass via NET in shared memory case
> > - set  vio_pipe_read_timeout in named pipe case.
> 
> actually only two, noone should call setsockopt directly, but use
> 'vio_timeout' which calls the function pointer, The function installed
> at the function pointer for sockets, uses setsockopt. At least that's
> how it's supposed to be.
> 
> I guess NET would be ok since it's already an alternative, but why not
> fix that as well? :)

Yeah, I fixed it in the last commit http://lists.mysql.com/commits/88636 .
This  expands the scope of the patch somewhat, but well, the shared memory's
logic ignoring vio_timeout, reading from the NET structure when appropriate
was not the best idea from the start on, I agree. I removed the NET from VIO
and added 2 integers, used by both shared memory and pipes.


Thread
bzr commit into mysql-5.0 branch (vvaintroub:2829) Bug#31621 Bug#47571Vladislav Vaintroub29 Oct
  • Re: bzr commit into mysql-5.0 branch (vvaintroub:2829) Bug#31621Bug#47571Magnus Blaudd30 Oct
    • RE: bzr commit into mysql-5.0 branch (vvaintroub:2829) Bug#31621Bug#47571Vladislav Vaintroub30 Oct
      • Re: bzr commit into mysql-5.0 branch (vvaintroub:2829) Bug#31621Bug#47571Magnus Blaudd30 Oct
        • RE: bzr commit into mysql-5.0 branch (vvaintroub:2829) Bug#31621Bug#47571Vladislav Vaintroub30 Oct