Re: patch 2.5.4 ISO C90 compliance patch

[Paul Edwards]

> Thanks for the patch.  Some comments:
> 
> I don't understand the first and third hunks of the proposed change to
> patch/patch.c; they seem irrelevant to the issue of porting 'patch'.

It is not possible to rename a file on MVS via rename().
You need to use a utility to do that.  You don't do it by
giving it a couple of DD names.  The rename() will simply
return failure, as it is allowed to do.

Since the user has already specified the filename for
the reject file, that should be used directly rather than
going via a temporary file.

> I don't understand the second hunk, either.  The emulation package
> has fdopen, so why can't it be used here, too?

Let me revisit that.  I thought it was not possible to emulate
that, but maybe it is.  The emulation package was not meant
to provide full functionality.

> I assume you fixed the strncasecmp-is-missing bug by using gnulib's
> version; that's what I was planning to do.

I used the version from diffutils.

> From our recent diffutils conversation it appears that changes like
> this:
> 
>    + #if HAVE_SYS_TYPES_H
>      #include <sys/types.h>
>    + #endif
> 
> won't be needed, as you can arrange for '#include <sys/types.h>' to be
> a no-op on your platform by using the appropriate DD statement.

My patch is to make the code C90 compliant.  People who
want a C90 compliant patch can search for it in bug-gnu-utils,
regardless of whether it is actually accepted into the baseline.

I intend to use the patched "patch 2.5.4" rather than a kludged
CRT myself, but others can do as they wish.

BFN.  Paul.