Re: emacsclient horribly broken

[Juanma Barranquero]

On 10/31/06, Tim Van Holder <address@hidden> wrote:
> My daily build from CVS currently fails because of (apparently
> Windows-centric) changes to lib-src/emacsclient.c.

Windows-based, not Windows-centric. I mean: I was doing the work on
Windows, but I tried to be compatible. Obviously I failed, though I'd
hesitate to call "horribly broken" to something you've been able to
fix with five lines of code. :)

> - a '}' outside of an #if/#endif block

Yeah, I moved this code around and left the "}" at the wrong place. Sorry.

> - <netinet/in.h> not included => lots of missing types and macros

I don't have a GNU/Linux at hand, so I cannot test this kind of thing.
Help much appreciated.

> - use of non-existent INVALID_SOCKET macro

My bad. I trusted Microsoft. I shouldn't.


> - ioctl(...FIONBIO...) used instead of the posix
>   fcntl(...O_NONBLOCK...)

I try to be compatible between GNU/Linux, Windows, etc. Being POSIX is
not a goal.

> + no headers/configury used for it; if
>   FIONBIO is actually desirable, configure should be adjusted to
>   check for <sys/filio.h> and <sys/ioctl.h>, so emacsclient can
>   include the ones that exist

Please, send a patch for this.

> - 'argv[0]' used instead of 'progname' (this at least should also
>   cause a compiler error on win32)

No, it doesn't (do you think I didn't compile it on Emacs before
commiting the change?). It's in a piece of code that get's #ifdef'd
out on Windows (because all the Unix socket code gets ignored).

> Patch follows (may not be complete solution but at least allows
> compilation to pass on my i386-pc-linux-gnu system).

I've installed it. Thanks!

--
                   /L/e/k/t/u