diff options
author | Kevin Ryde <user42@zip.com.au> | 2003-01-04 22:37:37 +0100 |
---|---|---|
committer | Kevin Ryde <user42@zip.com.au> | 2003-01-04 22:37:37 +0100 |
commit | 66477cc36191378991ec990d89a1d0c632527072 (patch) | |
tree | 24e3b45cc9152c8dec63511c306b1fad6bbe6f80 /acinclude.m4 | |
parent | 24bcd4fc33786b6a7f1ec8dc55477545ae15ae1f (diff) | |
download | gmp-66477cc36191378991ec990d89a1d0c632527072.tar.gz |
* acinclude.m4 (GMP_PROG_NM): Add some notes about failures, per
report by Krzysztof Kozminski.
Diffstat (limited to 'acinclude.m4')
-rw-r--r-- | acinclude.m4 | 10 |
1 files changed, 10 insertions, 0 deletions
diff --git a/acinclude.m4 b/acinclude.m4 index 6f597bd11..e4269e237 100644 --- a/acinclude.m4 +++ b/acinclude.m4 @@ -332,6 +332,16 @@ dnl $NM set to "nm", so $NM can't be assumed to actually work. dnl dnl A user-selected $NM is always left unchanged. AC_PROG_NM is still run dnl to get the "checking" message printed though. +dnl +dnl Perhaps it'd be worthwhile checking that nm works, by running it on an +dnl actual object file. For instance on sparcv9 solaris old versions of +dnl GNU nm don't recognise 64-bit objects. Checking would give a better +dnl error message than just a failure in later tests like GMP_ASM_W32 etc. +dnl +dnl On the other hand it's not really normal autoconf practice to take too +dnl much trouble over detecting a broken set of tools. And libtool doesn't +dnl do anything at all for say ranlib or strip. So for now we're inclined +dnl to just demand that the user provides a coherent environment. AC_DEFUN(GMP_PROG_NM, [dnl Make sure we're the first to call AC_PROG_NM, so our extra flags are |