[OpenAFS-devel] Compiling openafs cvs checkout on dux_51 (Was: RE: [OpenAFS-devel] Testers needed for Prototypes cvs branch ofopenafs)

[Mattias Amnefelt]

Martin MOKREJŠ <mmokrejs@natur.cuni.cz> writes:

Is there any reason why you're not simply using systype alpha_dux50? It
compiles for me on any 5.x.

> > You'll want to call flex with -l on OSF to get yylineno defined.
> 
> OK, editing config.status to supply "flex -l" instead of flex and
> rerunning it helps, so here's one error, which I "fixed" myself ....

You'll get this with systype alpha_dux50. This is defined in src/cf/osconf.m4

[snip]

> Commenting out both defines as below in src/lwp/lwp.c:
> 
> #ifdef  AFS_OSF_ENV
> /*extern void *malloc(int size);
> extern void *realloc(void *ptr, int size);*/
> #endif

> pushes me further, but at the end another error:

[snip]

> cc: Error: ./fs_conv_411.c, line 487: In this statement, "umount" expects 2 arguments, but 1 are supplied. (toofewargs)
>        umount(path);
> -------^

You get conflicting types because you don't have _NO_PROTO defined. This is
also defined in src/cf/osconf.m4

It's debatable whether it's a good thing to have to define, or not, but
that's the way the buildsystem works.

In short, I'd recomend you to use alpha_dux50 for systype instead.

Also, for the flock thing in your last email. I don't think we're affected
by flocks not being inherited across forks. From what I've seen we don't
depend on a child getting an flocked file from it's parent. 

Others might still be interested to read
http://www.xray.mpe.mpg.de/mailing-lists/tru64-unix-managers/2002-04/msg00435.html

/mattiasa