[Axiom-developer] Debian/Woody Build failure even with Gawk 3.1.4 [ long mail ]

[Balbir Thomas]

Hi,
The old problem of stray backslashes in src\algebra\Makefile
still occurs with the Debian testing/unstable version
of gawk (version 3.1.4) used on Debian/Stable platform.
The make version is 3.79.1. Make version on Debian/testing
stable is 3.80. Hence it is still not clear to me why
axiom can be built on Debian testing/unstable but not
on stable. There are no apparent build dependencies that
can not be fulfilled on Stable. 

Here is the same error message :

make[4]: *** No rule to make target
`/home/bt/archive/axiom-20050201/int/algebra/ABELGRP.o', needed by
`src'.  Stop.
make[4]: Leaving directory `/home/bt/archive/axiom-20050201/src/algebra'
make[3]: *** [algebradir] Error 2
make[3]: Leaving directory `/home/bt/archive/axiom-20050201/src'
make[2]: *** [srcdir] Error 2
make[2]: Leaving directory `/home/bt/archive/axiom-20050201'
make[1]: *** [all] Error 2
make[1]: Leaving directory `/home/bt/archive/axiom-20050201'
make: *** [build-stamp] Error 2

Here is some sample lines from src/algebra/Makefile

\${MID}/ABELGRP.spad: \${IN}/catdef.spad.pamphlet
        \${TANGLE} -R"category ABELGRP AbelianGroup" 
\${IN}/catdef.spad.pamphlet>\${MID}/ABELGRP.spad

\${MID}/ABELGRP.lsp: \${IN}/catdef.spad.pamphlet
        \${TANGLE} -R"ABELGRP.lsp BOOTSTRAP" 
\${IN}/catdef.spad.pamphlet>\${MID}/ABELGRP.lsp

\${MID}/ABELGRP-.lsp: \${IN}/catdef.spad.pamphlet
        \${TANGLE} -R"ABELGRP-.lsp BOOTSTRAP" 
\${IN}/catdef.spad.pamphlet>\${MID}/ABELGRP-.lsp


Also some very unusual behaviour with the build process when I do edit
src/algebra/Makefile to remove all backslashes before dollar signs.
On doing so the build can be restarted and all of algebra is built 
even the tests are run but then (as pointed out before) build fails 
with and error message :

BOOT>cp: cannot create regular file
`/home/bt/archive/axiom/mnt/linux/algebra/RETRACT-.NRLIB/code.o': No
such file or directory
make[4]: *** [/home/bt/archive/axiom/int/algebra/RETRACT-.NRLIB/code.o]
Error 1

The strange thing is that if I restart the build it continues one step
further every time I restart it. 

For instance after staring the build two or there time it now fails at

BOOT>cp: cannot create regular file
`/home/bt/archive/axiom/mnt/linux/algebra/ELTAGG-.NRLIB/code.o': No such
file or directory
make[4]: *** [/home/bt/archive/axiom/int/algebra/ELTAGG-.NRLIB/code.o]
Error 1

Then at :

BOOT>cp: cannot create regular file
`/home/bt/archive/axiom/mnt/linux/algebra/IEVALAB-.NRLIB/code.o': No
such file or directory
make[4]: *** [/home/bt/archive/axiom/int/algebra/IEVALAB-.NRLIB/code.o]
Error 1

and so on.

Now the build rule for this step is (if I understand correctly)

${OUT}/%.o: ${MID}/%.NRLIB/code.o
        @ echo generic 1 -- making ${OUT}/$*.o from ${MID}/$*.NRLIB
        @ cp ${MID}/$*.NRLIB/code.o ${OUT}/$*.o

where the variables are being defined as 
MID=${INT}/algebra
OUT=${MNT}/${SYS}/algebra

This means $* in the copy statement was translated as
RETRACT-.NRLIB/code !!!. But when the build was  restarted it did check
for the RETRACT.o in ${OUT} which was there from before (?) and so it 
went on to the next step. 

This is weird as hell :-) ! I doubt this is some serious corruption of
my linux distribution as I generally try to keep a clean Debian/Stable
distribution. Sould someone like to inspect I can put the whole
directory tree of the current state of axiom build online on my
homepage.

I'll continue working on this as time permits and see if I can figure out 
anything else.
I hope you will kindly bear with the nuisance of me posting about my successes 
or
failures :-). Though if some axiom developer is trying to build on
Debian/Stable I'll stop, as I am sure he/she is more able than me to
figure this out.

sincerely
b thomas