Re: Do not report build errors in bugzilla

[prj at po dot cwru dot edu (Paul Jarc)]

Roland McGrath <roland@redhat.com> wrote:
> Would you like to figure out what the exact requirements are and
> help us document them?

I don't have time to coordinate that effort, but I can contribute data
points as I find them.

> However, if you fix it yourself and send us clean patches that make
> glibc's build work with any POSIX.2-compliant sh, we will certainly
> put those in.

SUSv3 doesn't specify -e or -E for echo, and csu/Makefile uses -e, so
it's already less portable than intended, I guess.  Here's a slightly
different patch for my case that uses -e instead of -E, though, so at
least it won't reduce portability any further; I'm testing it now:

	* csu/Makefile: Use "echo -e" whenever we care how backslash
	is handled.

Index: csu/Makefile
===================================================================
RCS file: /cvs/glibc/libc/csu/Makefile,v
retrieving revision 1.73
diff -u -r1.73 Makefile
--- csu/Makefile        8 Mar 2004 21:28:15 -0000       1.73
+++ csu/Makefile        17 Aug 2004 20:14:06 -0000
@@ -231,13 +231,13 @@
                   if [ -z "$$os" ]; then \
                     os=Linux; \
                   fi; \
-                  echo "\"Compiled on a $$os $$version system" \
-                       "on `date +%Y-%m-%d`.\\n\"" ;; \
+                  echo -e "\"Compiled on a $$os $$version system" \
+                       "on `date +%Y-%m-%d`.\\\\n\"" ;; \
           *) ;; \
         esac; \
         files="$(all-Banner-files)";                           \
         if test -n "$$files"; then                             \
-          echo "\"Available extensions:\\n\"";                 \
+          echo -e "\"Available extensions:\\\\n\"";            \
           sed -e '/^#/d' -e 's/^[[:space:]]*/  /'              \
               -e 's/^\(.*\)$$/\"\1\\n\"/' $$files;             \
         fi) > $@T


>> There's also the __i686 macro defined by gcc
...
> In fact, I don't know what you are talking about.

Passing -march=i686 makes gcc implicitly define __i686 to 1, but some
of glibc's assembly code uses __i686 without expecting it to be a
macro.  I think this is the least disruptive change to handle it:

http://sources.redhat.com/ml/libc-alpha/2002-10/msg00157.html

More threads:
http://sources.redhat.com/ml/libc-alpha/2003-05/msg00125.html
http://sources.redhat.com/ml/libc-alpha/2003-01/msg00119.html


paul