Re: Patch: rename to `insight'

[Tom Tromey <tromey at cygnus dot com>]

>>>>> "Eli" == Eli Zaretskii <eliz@delorie.com> writes:

Eli> I would suggest to consider a short C program instead of a shell
Eli> script, for portability's sake.  Shell scripts are not very
Eli> portable to non-Posix platforms.

I assume you're talking about Windows.  I thought gdb required Cygwin
to run on Windows, in which case you will have a shell.  Or is there a
real native gdb?

>> +	$(SHELL) $(srcdir)/../mkinstalldirs $(bindir)

Eli> Shouldn't this say "-$(SHELL)", in case $(bindir) already exists?

No.  mkinstalldirs exits with status 0 if the directory already
exists.

Eli> This assumes that the distribution is not moved once installed.
Eli> That might be true on a typical Unix system (even there you might
Eli> have problems with CD-ROM distributions), but it will make lots
Eli> of problems on DOS/Windows systems, where a precompiled binary is
Eli> usually installed into a directory that has nothing to do with
Eli> the original installation at "make install" time.

The program first tries to see if `gdb' exists in the same directory
as `insight'.  Most shells use the full path as $0, so this has a
good chance of working.

Eli> Why not simply invoke "gdb" and let PATH do its thing?  Or at
Eli> least try that before falling back on the installation directory?

The installation directory should be the first choice (after the
directory of `insight' itself).

Using PATH isn't so good because people sometimes use the full path to
run something that isn't in their PATH.  This happens a lot with
developers.  An earlier version of this patch didn't use bindir -- and
I ran into this problem myself (I have 3 or 4 versions of gdb around
at any given time).

I fixed the script to fall back on PATH if it can't find the file any
other way.

Eli> I'm not sure this will DTRT with all DOS/Windows ports of Unixy
Eli> shells.  I suggest to use "gdb.exe" on those platfrms (better
Eli> safe than sorry).

Ok.  I added $(EXEEXT) to the @gdb@ substitution.

Eli> Writing a short C program would eliminate all these problems (and
Eli> also avoid requiring users to have a ported shell on their
Eli> machines).  That's why I suggested that in the first place.

I don't know that much about Windows or DOS.  I don't know how I would
determine whether the program was portable there.

Tom