Re: non-blocking stdin on windows

[Eric Covener <co...@gmail.com>]

On 9/27/07, Eric Covener <co...@gmail.com> wrote:
> httpd's mod_ext_filter sets up a child process with a stdin of
> APR_CHILD_BLOCK, but on windows the parent side of the stdin pipe
> defaults to a "non-blocking pipe with an infinite (-1) timeout".  Such
> a pipe in unix has its apr_file_pipe_timeout() flipped to 0 during
> apr_procattr_io_set().
>
> Net, a subsequent apr_file_read() behaves differently on the two
> platforms when no data is available w/o the explicit
> apr_file_pipe_timeout_set(foo, 0) call.
>
> While the API might be a little ambiguous, and the caller can
> explicitly set the timeout, is this a discrepancy APR should
> eliminate?

I'm going to add the apr_file_pipe_timeout_set(foo, 0) call instead to
mod_ext_filter unless there are any objections.

-- 
Eric Covener
covener@gmail.com

Re: non-blocking stdin on windows

["William A. Rowe, Jr." <wr...@rowe-clan.net>]

Eric Covener wrote:
>>
>> While the API might be a little ambiguous, and the caller can
>> explicitly set the timeout, is this a discrepancy APR should
>> eliminate?
> 
> I'm going to add the apr_file_pipe_timeout_set(foo, 0) call instead to
> mod_ext_filter unless there are any objections.

+1

Re: non-blocking stdin on windows

["William A. Rowe, Jr." <wr...@rowe-clan.net>]

Eric Covener wrote:
>>
>> While the API might be a little ambiguous, and the caller can
>> explicitly set the timeout, is this a discrepancy APR should
>> eliminate?
> 
> I'm going to add the apr_file_pipe_timeout_set(foo, 0) call instead to
> mod_ext_filter unless there are any objections.

+1