httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jim Jagielski <>
Subject Re: cvs commit: apachen/src/test time-sem.c
Date Thu, 22 Jan 1998 23:52:47 GMT
Marc Slemko wrote:
> On 22 Jan 1998 wrote:
> > jim         98/01/22 15:12:10
> > 
> >   Modified:    .        STATUS
> >                htdocs/manual/misc FAQ.html perf-tuning.html
> >                htdocs/manual/mod core.html
> >                src      CHANGES PORTING
> >                src/main conf.h http_main.c httpd.h
> >                src/os/win32 os.h
> >                src/test time-sem.c
> >   Log:
> >   Submitted by:	Jim Jagielski
> >   
> >   Do make the code a bit clearer, some minor #define changes (and
> >   the resultant flow-thru in the docs).
> >   
> >      HAVE_MMAP                 ->   USE_MMAP_SCOREBOARD
> >      HAVE_SHMGET               ->   USE_SHMGET_SCOREBOARD
> Would it be better to keep HAVE_MMAP and HAVE_SHMGET and add defines for
> what to use for the scoreboard, that default to something depending on
> what HAVE_ are defined yet can be overriden in the OS-specific parts or
> elsewhere?
> There _is_ a use for HAVE_MMAP, etc. and could be more uses in the
> future.  Consider adding a generic shared memory interface, then you
> need yet more defines and they are all based on seperate things that
> all have to be set for each platform.

Yep... this is part of what I have in mind. Beforehand, we had HAVE_MMAP
and HAVE_SHMGET but not in a logical way (for example, you could
have HAVE_MMAP not defined, but still use USE_MMAP_FILES).

Right now, we don't have any sections of code that use HAVE_MMAP
or HAVE_SHMGET unless we did something ugly but logical like

 #if defined(HAVE_MMAP) && defined(USE_MMAP_SCOREBOARD)

but that seemed overkill for now. It _is_ good to know if we have
mmap() and/or shmget() but right now it would result in either
#def's that weren't used, or the above time of preprocessor lines :/
Neither option seemed too nice. I didn't see any outside modules
that required them either, but most likely that may not be universally
true, in which case we fold them back in, but now logically :)

For 2.0, then I foresee lots of the above type of codelets

      Jim Jagielski            |       jaguNET Access Services           |
            "Look at me! I'm wearing a cardboard belt!"

View raw message