Sender: bkorb AT sco DOT COM
Message-ID: <397EEC40.9177C526@sco.com>
Date: Wed, 26 Jul 2000 06:48:48 -0700
From: Bruce Korb <bkorb AT sco DOT COM>
Organization: Santa Cruz Operations
X-Mailer: Mozilla 4.7 [en] (X11; I; SCO_SV 3.2 i386)
X-Accept-Language: en
MIME-Version: 1.0
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
CC: djgpp-workers AT delorie DOT com
Subject: Re: DJGPP patch for fixincludes
References: <Pine DOT SUN DOT 3 DOT 91 DOT 1000726093336 DOT 3890B-100000 AT is>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com

Eli Zaretskii wrote:
> Were these diffs generated against the current CVS, or to the last
> released version (or something else)?

CVS, but I'll be applying the first patch today or tomorrow.

> > All fix tests that depend on using a server shell are presumed
> > to NOT APPLY in DOS-land.  All fixes that use a shell script
> > to produce the modification (all of one example anyway) are
> > also presumed to NOT APPLY.
> 
> Hmm... I'm afraid I don't know enough about this to comment off the
> top of my head.  DJGPP does have a working port of Bash, and our
> system(3) knows how to run a script if Bash is installed, so if the
> fixes performed by a shell script are useful, we could retain them.

"if Bash is installed".  I assure you they are not relevant to DOS
anyway:

  /*
   *  sys/lc_core.h on some versions of OSF1/4.x pollutes the namespace by
   *  defining regex.h related types.  This causes libg++ build and usage
   *  failures.  Fixing this correctly requires checking and modifying 3 files.
   */
  fix = {
    hackname = osf_namespace_a;
    files    = reg_types.h;
    files    = sys/lc_core.h;
    test     = " -r reg_types.h";
    test     = " -r sys/lc_core.h";
    test     = " -n \"`grep '} regex_t;' reg_types.h`\"";
    test     = " -z \"`grep __regex_t regex.h`\"";
    ...

  fix = {
    hackname = uw7_byteorder_fix;
    files    = arpa/inet.h;
    select   = "in_port_t";
    test     = "-f $DESTDIR/sys/byteorder.h";
    ...

  fix = {
    hackname = vxworks_needs_vxworks;
    files    = sys/stat.h;
    test     = " -r types/vxTypesOld.h";
    test     = " -n \"`egrep '#include' $file`\"";
    test     = " -n \"`egrep ULONG $file`\"";
    select   = "#[ \t]define[ \t]+__INCstath";
    ...

  fix = {
    hackname = vxworks_time;
    files    = time.h;
    test     = " -r vxWorks.h";
    ...

> I have a few comments, based on code inspection; hopefully, they will
> be useful to whoever works on this (or to myself ;-).
> 
> > +   pz_tmptmp = (char*)xmalloc( strlen( argv[4] ) + 5 );
> > +   sprintf( pz_tmptmp, "%sX", argv[4] );
> > +   if (freopen( pz_tmptmp, "w", stdout ) != stdin)
>                                                ^^^^^
> Shouldn't this be stdout?

:-)

> > + # ifdef __DOS__
> > +   {
> > +     tSCC z_tmp_fname_fmt[] = "%s/fxinc%03X.tp";
> > +     char* pz = getenv( "TMP" );
> > +     if (pz == NULL)
> > +       pz = "/tmp";
> > +     pz_temp_file = (char*)xmalloc (sizeof(z_tmp_fname_fmt) + strlen (pz));
> > +     sprintf (pz_temp_file, z_tmp_fname_fmt, pz, getpid() & 0x0FFF);
> > +   }
> > + # endif
> 
> This should probably use $TMPDIR or tempnam().

Yes.  This:

> > + # ifdef __DOS__
> > +   pz_temp_file =  tempnam( "/tmp","fxinc");
> > + # endif

is much simpler.

> > ! #else
> > ! /*
> > !  *  IF we are in MS-DOS land, then whatever shell-type test is required
> > !  *  will, by definition, fail
> > !  */
> > ! #define test_test(t,tf)  SKIP_FIX
> > ! #endif
> 
> See my comment above: if Bash is installed (and it must be, in order
> to build GCC), these shell tests need not fail.
> 
> Thanks again.

Well, all of the existing tests are irrelevant, but if you want a proper
fix, then, yes, run popen() and read back the true or false result.