Re: libtool woes

[Jukka Santala <jsantala@tml.hut.fi>]

On Mon, 21 Jan 2002, Alexandre Petit-Bianco wrote:
> Jukka Santala writes:
> > The fix in gnu.readline is neccessary because otherwise the build
> > will break horribly on any systems with a sensible jar-tool that
> > actually requires source-filenames.
> What are you seeing going wrong with the original actions?

First, a funny, using GCC 3.1/HEAD jar:

(here=`pwd`; cd ./upstream/src; jar uf $here/org-gnu-readline-0.6.jar)
jar: `-u' mode unimplemented.
make: *** [org-gnu-readline-0.6.jar] Error 1

In other words, the jar utility provided with GCC doesn't support -u at
all, hence some rewording might be in order.

Using Sun j2sdk 1.3.1's or IBM's 2-13 jar:
(here=`pwd`; cd ./upstream/src; jar uf $here/org-gnu-readline-0.6.jar)
'u' flag requires manifest or input files be specified!
Usage: jar {ctxu}[vfm0M] [jar-file] [manifest-file] [-C dir] files ...

Kaffe's jar does work, but that's about it. The main reason is that the
line doesn't actually do anything, since there are no non-source-files to
add. To make it able to compile with the present GCC alone, the additional
resource-files need to be added to the above "jar cf" line ofcourse.

> > Index: gnu.regexp/Makefile.am
> > Index: xerces/Makefile.am
> This is fine. Could you please write ChangeLog entries?

Sure. These do not allow a cross-compile yet, by the way, but they're
likely to bite even people who're just trying out alternative compiles of
GCC, and given rhug doesn't yet compile with the stable/release version,
that's quite a large portion of them :) Same problem goes with gcjh etc.
though, so I guess I'll look if I can bundle more things into that patch
at once.

 -Jukka Santala