[RFC] server/util_fcgi.c

[Jeff Trawick <tr...@gmail.com>]

I'm planning to add FastCGI structures and a few support routines to core
for (initially) sharing between mod_authnz_fcgi and mod_proxy_fcgi.  The
size is pretty small so <em>to me it seems that</em> the overhead of
creating some hypothetical mod_fcgi_util and having uglier APIs (optional
function load, handling error path for missing module, etc.) doesn't seem
warranted.  But YMMV, so please follow up if you disagree.

Attached is the patch to give you an idea of what I'm talking about.  It is
largely lifted from mod_proxy_fcgi, though it doesn't look much different
from mod_fcgid or the spec or presumably any other implementation.  Some of
the code from mod_proxy_fcgi has been made more general (e.g., separating
building of protocol data from I/O).

Planned changes before commit:

* Fix ap_fcgi_encoded_env_len()/ap_fcgi_encode_env() so that it is
straightforward to process more of the environment table after sending some
of the environment in a FastCGI record.  (Currently mod_proxy_fcgi
truncates once it reaches the limit for a single FastCGI record, but we
should be able to send over a larger env using multiple records.  What is
in the patch was an incomplete first attempt.)
* Windows and other platform build changes
* See if symbol names need changing for brevity or readability

-- 
Born in Roswell... married an alien...
http://emptyhammock.com/

Re: [RFC] server/util_fcgi.c

[Jeff Trawick <tr...@gmail.com>]

On Sun, Aug 11, 2013 at 8:17 AM, Jeff Trawick <tr...@gmail.com> wrote:

> I'm planning to add FastCGI structures and a few support routines to core
> for (initially) sharing between mod_authnz_fcgi and mod_proxy_fcgi.  The
> size is pretty small so <em>to me it seems that</em> the overhead of
> creating some hypothetical mod_fcgi_util and having uglier APIs (optional
> function load, handling error path for missing module, etc.) doesn't seem
> warranted.  But YMMV, so please follow up if you disagree.
>
> Attached is the patch to give you an idea of what I'm talking about.  It
> is largely lifted from mod_proxy_fcgi, though it doesn't look much
> different from mod_fcgid or the spec or presumably any other
> implementation.  Some of the code from mod_proxy_fcgi has been made more
> general (e.g., separating building of protocol data from I/O).
>
> Planned changes before commit:
>
> * Fix ap_fcgi_encoded_env_len()/ap_fcgi_encode_env() so that it is
> straightforward to process more of the environment table after sending some
> of the environment in a FastCGI record.  (Currently mod_proxy_fcgi
> truncates once it reaches the limit for a single FastCGI record, but we
> should be able to send over a larger env using multiple records.  What is
> in the patch was an incomplete first attempt.)
> * Windows and other platform build changes
> * See if symbol names need changing for brevity or readability
>

and actually change MODULE_MAGIC_NUMBER_MINOR (what is it about that task?)

-- 
Born in Roswell... married an alien...
http://emptyhammock.com/