Re: cvs commit: apache/src multithread.h

[Randy Terbush <ra...@zyzzyva.com>]

> Randy Terbush wrote:
> >     #define APACHE_TLS
> >     /* Only define the ones actually used, for now */
> >     #define create_mutex(name)	NULL
> >   ! #define acquire_mutex(mutex_id)	{}
> >   ! #define release_mutex(mutex_id)	{}
> 
> The snag with doing this is if people start testing the return values. It'll do
> for now, though.
> 
> Cheers,
> 
> Ben.

Yeah... I even went back to this several times trying to find a way 
to satisfy the compiler so that the code could in fact return a 
value. The only true solution I see is to create the stub functions 
for for the UNIX systems.

How about if we create a /src/os directory and put the nt stuff in
/src/os/nt and start a /src/os/unix for things like this?

Re: cvs commit: apache/src multithread.h

[Paul Sutton <pa...@ukweb.com>]

On Mon, 16 Jun 1997, Randy Terbush wrote:
> How about if we create a /src/os directory and put the nt stuff in
> /src/os/nt and start a /src/os/unix for things like this?

Should the os directory exist at the top level instead of inside src? We
have compiled programs in /support and (possibily in the future) /cgi-src
which will need access to the same multi-platform abstractions.

Also there should be an emx directory as well as nt and unix, so we can
get rid of the ifdef __EMX__'s throughout the code.

Umm, are all the unix variants going to be treated as one platform as far
as this abstraction is concerned? It might be neater to have a directory
per Unix variant as well, to stop the conf.h getting longer and longer. 

//pcs