Re: [PATCHes] gawk-3.1.0 build problems under DJGPP

[Peter J. Farley III]

At 11:51 AM 8/7/01 -0700, Scott Deifik wrote:
>Greetings. Re this:
<Snipped>
>     Thank you for your patches.  Traditionally, it has been hard to
>compile and test gawk in DOS environments.  For that reason, we have
>maintained two separate files: pc/makefile for building and
>pc/makefile.tst for testing.  Together, these work for both DJGPP &
>Microsoft C v8 (among other compilers).  For instance, in
>pc/makefile.tst, CMP defaults to using diff, as you suggest.

Yes, I did see those files.  With the diligence and zest of the DJGPP 
maintainers, however, it *has* been getting easier and easier to 
compile and test GNU utilities on DOS.

>     I'm personally in favor of making the main distribution able to
>handle DOS-specific needs (meaning various DOS-based compilers and
>environments) if the changes would serve all/most environments.  The
>only question is what to do when DOS has it's own way of doing 
things.
>At that point, the decision of whether or not to specifically test 
for
>the different need is a judgement call.

I agree about making *ix-based distributions DTRT for as many 
environments as possible, without compromising the distribution.  It 
certainly is a judgement call where that line is.

In this case, I think the "|&" internet capabilities are the "hard 
problem", in the mathematical sense, for DJGPP/MS-DOS due to the normal 
lack of IP communication stacks.  For that reason, the changes I have 
suggested to ignore the "clos1way" test would have to be pretty much 
permanent until/unless such a stack is normally available.

The "pid/ppid" differences might be resolvable, might not be.  The 
current DJGPP implementation doesn't support what *ix systems deliver, 
so that the change I suggested to bypass this test is likely to be 
relatively long-lived.  However, there is an active DJGPP maintainer 
team, and it is certainly possible that a solution could be found, 
though perhaps not for all cases.

I feel that the other changes I suggested should be permanent in the 
sense that they DTRT, at least as far as my limited experience can 
tell.  I will certainly accept the community's (and the prime 
implementor's) judgement on them, however, since my experience in these 
things is limited.

>     I'm not sure what the final determination will be about each of
>your (good) suggestions, but I just wanted to thank you for your 
input
>and let you know about pc/makefile.tst (in case you were unaware).

First of all, you are quite welcome.  I try to contribute what I can, 
when I can.

I did start with the pc/* files, but wanted to see if the package could 
be built as it was intended to be built.  Plus, the "pre-built" files 
*assume* a location for the "root" of the DJGPP environment, which I 
have had several occasions to change on my system.  Using the v2.03 
DJGPP notation "/dev/env/DJDIR" makes the files that are built 
independent of the "root" location, and configure with argument 
"--prefix=/dev/env/DJDIR" seemed to me to be the "right way" to 
accomplish that goal.  Plus it represented another opportunity for me 
to continue my education in configure and make and the tools used to 
build them.

However, I should also make clear that "stand-alone" DOS is not one of 
the environments I am concerned with, so I have luxuries that the DJGPP 
maintainers do not have, like not worrying about 8.3 filename 
restrictions.

---------------------------------------------------------
Peter J. Farley III (address@hidden)