Date: Sun, 7 Sep 1997 10:44:21 -0400 (EDT)
Message-Id: <199709071444.KAA21429@delorie.com>
From: DJ Delorie <dj AT delorie DOT com>
To: pjfarley AT dorsai DOT org
CC: djgpp AT delorie DOT com
In-reply-to: <3410b47f.1231392@snews.zippo.com> (pjfarley@dorsai.org)
Subject: Re: Rebuilding gcc -- errors in tests (djtst201.zip)
Precedence: bulk


> 1.  "gcc" is used to compile *.cc source without the gxx options, and
> compile errors result due to no libg++ includes available.

gcc knows how to compile *.cc files without extra options.  Perhaps
you didn't download the C++ libraries and headers?

> 2.  The following error is fairly self-explanatory:
> 
> make.exe: *** No rule to make target `/v1/lib/libc.a', needed by
> `oldlibc.arm'.
> 
> Obviously, those who have only downloaded v2 do not have a
> "/v1/lib/libc.a" or "/v1/lib/libpc.a" or "/v1/lib/libm.a", all of
> which are needed to make "oldlibc.arm".

That is only for me, so that I can make sure v2 has at least the same
functionality as v1.  Note that the next version will no longer have
this test.

> 3.  There do not seem to be any test executions of the .exe files that
> are built in the test subdirectories, nor sample output nor
> instructions for running them manually.

The test programs are for libc developers, so that they have something
to test with when they modify libc.  It is not an automatic regression
test.

> Q1:  Is the compare of "newlibc.arm" and "oldlibc.arm" significant or
> necessary to the tests?

No.  The other half of that makefile *is* critical.  It guarantees
that we don't violate the ANSI or POSIX namespace rules.

> Q2:  Is the success in building the .exe files test enough, or is
> there more that should be done to confirm correct operation?

There is no regression suite for DJGPP.  The tests there are examples
for libc developers, to assist them in testing the functions.

It would be good to have a real automatic regression suite, but we
don't.