Re: rcs-5.8-1 on HP-UX 11.31

On Mon, Jun 25, 2012 at 01:44:17PM +0200, Thien-Thi Nguyen wrote:
> () Paul Ackersviller <[email protected]>
> () Mon, 25 Jun 2012 03:05:43 +0000
> 
>    On Wed, Jun 06, 2012 at 11:19:25AM +0200, Thien-Thi Nguyen wrote:
>    > () Paul Ackersviller <[email protected]>
>    > () Mon, 23 Apr 2012 02:48:33 +0000
>    > 
>    > I have just released 5.8.1.  Could you please try it out and let
>    > me know if these problems persist?
> 
>    I didn't try quite exactly the 5.8.1 release, but the git head
>    from a few days ago.  That should be close enough, right?
> 
> If you mean branch ???master???, then no, that is somewhat beyond 5.8.1
> (added features (and bugs)).  Try branch ???a??? or tag ???5.8.1??? instead,
> which corresponds directly.

Alright, I took the 5.8.1 tag and got all the same results -- everything
compiles and the tests all pass.

> Thanks.  I've briefly looked at the (non gnulib) warnings, and
> identified two legitimate errors.  One (use before init) in b-eph.c
> is normally masked for me because ('/' != SLASH) is never true, but
> i would think that the same applies to an HPUX system and so wonder
> what i'm missing.

Looks to me like HP is being a little extra helpful in giving warnings even
for code that can be compile-time optimized out.  Would the non-initialization
only run on MS-DOS, or does that include modern Windows too?

>    One other thing, in case you care, I also ran the tests with memory
>    leak checking.  Following is the report for whatever it's worth.
> 
>    -------------------------------------------------------------------------
>    4064 bytes leaked at 0x40aa40a0 (98.55% of all bytes leaked)
>    #0  allocate() at b-divvy.c:49
>    #1  xmalloc() at b-divvy.c:95
>    #2  _obstack_begin() at obstack.c:173
>    #3  make_space() at b-divvy.c:78
> 
>    -------------------------------------------------------------------------
>    44 bytes leaked at 0x4059c170 (1.07% of all bytes leaked)
>    #0  allocate() at b-divvy.c:49
>    #1  make_space() at b-divvy.c:76
>    #2  pairnames() at rcsfnms.c:340
>    #3  rlog_main() at rlog.c:893
> 
>    -------------------------------------------------------------------------
>    16 bytes leaked at 0x4005d220 (0.39% of all bytes leaked)
>    #0  allocate() at b-divvy.c:49
>    #1  make_space() at b-divvy.c:73
>    #2  pairnames() at rcsfnms.c:340
>    #3  rlog_main() at rlog.c:893
> 
>    -------------------------------------------------------------------------
> 
> Hmm, could it be possible that the obstacks RCS uses is confusing
> the memory leak checking program?  (What would that be, btw?)

I think it's called real-time-checking, but isn't a separate product
that I'm aware of, just bundled with HP's compiler.

One additional question I just remembered -- is it normal now to require
GNU make?  Most autoconf-generated builds traditionally suppored sysv
make as well.  Is this changing now?