egcs-980328 on i586-pc-linux-gnu build failure

egcs-980328 on i586-pc-linux-gnu build failure

[Horst von Brand]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1314 bytes --]

linux-2.1.91
binutils-2.8.1.0.27 (same result with binutils-2.8.1.0.25)
{C,CXX,F77}FLAGS = -O2 -fomit-frame-pointer -march=pentium
--enable-{shared,threads,haifa}

I get:

stage1/xgcc -Bstage1/ -c  -DIN_GCC   -W -Wall -O2 -fomit-frame-pointer -march=pentium  -W -Wall    -If -I./f -I. -I. -I./config f/expr.c -o f/expr.o
xgcc: Internal compiler error: program cc1 got fatal signal 6
make[3]: *** [f/expr.o] Error 1

Just to have something to run "make -k check" on, I built the third stage
by hand, got no differences in "make compare". To my surprise, the "make -k
check" went ahead and finished the build of g77 using gcc, not ".../xgcc
-B..." as a compiler.

So, a couple of observations:

- make -k bootstrap (at least) should propagate the -k down.  If this fails
  catastrophically (i.e., no cc1 built) it won't waste too much time either
  way ;-)

- There is no indication of how to build just some subset of the languages
  in the INSTALL instructions.

- make check should be smarter, and really rebuild (the g77 results that
  I'll get will be totally bogus)

I'm running make CC=".../xgcc -B..." -k check right now.  Will report
results soon.
-- 
Horst von Brand                             vonbrand@sleipnir.valparaiso.cl
Casilla 9G, Viña del Mar, Chile                               +56 32 672616

Re: egcs-980328 on i586-pc-linux-gnu build failure

[Jim Wilson]

	stage1/xgcc -Bstage1/ -c  -DIN_GCC   -W -Wall -O2 -fomit-frame-pointer -march=pentium  -W -Wall    -If -I./f -I. -I. -I./config f/expr.c -o f/expr.o
	xgcc: Internal compiler error: program cc1 got fatal signal 6
	make[3]: *** [f/expr.o] Error 1

This is the exact same problem as the regex.c failures, and I just checked in
a patch for that a little while ago.  So I believe this is fixed now.

	 To my surprise, the "make -k
	check" went ahead and finished the build of g77 using gcc, not ".../xgcc
	-B..." as a compiler.

Make check depends on make all, so yes this will do the wrong thing if make
bootstrap failed.  I don't see any obvious fix for it.  We can't make it
depend on make bootstrap, as this will fail for cross compilers.  I don't
think it is worth worrying about for the moment.

	- There is no indication of how to build just some subset of the languages
	  in the INSTALL instructions.

Because we don't provide a way to build only some of the existing languages.

You can avoid building a language by not ftp'ing it.  For instance if you
ftp only the core package, then you won't build the C++/Fortran front ends
because they won't exist.  But if you ftp the entire package, then you should
build all of the languages.

Disabling a language is not as easy as it was with FSF gcc2 release, because
we also have runtime libraries.  If you don't build C++, then you also have
to not build the libraries that contain C++ code, but we don't have Makefile
or configure support for that.

Jim