[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[GNU Crypto] Re: GCJ build
From: |
Olivier LF |
Subject: |
[GNU Crypto] Re: GCJ build |
Date: |
Thu, 24 Oct 2002 22:31:16 +1000 |
User-agent: |
Mutt/1.4i |
On Wed, Oct 23, 2002 at 11:05:33PM +1000, Raif S. Naffah wrote:
>
> that's a very good point! should we do the same for the top level
> directory too?
>
I'd say yes.
> doesn't the user need to have LIBTOOL anyway to use the gcj/ tools?
> doesn't including the Makefile.in as you point out in the previous part
> of the message enough to address the needs for people who want to just
> compile sources?
>
You need libtool for gcj but if you want to offer all types of builds
from the one "configure" script then you'll run into this problem.
Bytecode compilation does not need libtool or shared library plumbing,
so if I say:
configure --with-compiler=jikes --with-rt-jar=path/to/rt.jar
with a fully operational Jikes compiler, I expect it to work. In
practice it may fail on libtool or linker problems.
This is because "configure" must perform the AM_LIBTOOL stuff (and may
fail if thing are not to its liking). Also, it is of no use to
the Jikes compilation it must be performed!
The LIBTOOL step cannot be skiped otherwise the resulting Makefile
has syntax errors in it.
I can't remember the exact problem but from the top of my head you get
something like that:
Makefile.am includes the offending:
lib_LTLIBRARIES = lib-gnu-crypto.la
...
...
immediately, the generated Makefile.in includes targets like this one:
$(LIBTOOL_THIS): $(LIBTOOL_THAT)
$(AND_LIBTOOL_THAT) blah blah
"configure" resolves the variables in the AM_LIBTOOL step and produces
a Makefile similar to this:
this_libtool: that_libtool
and_more_libtool blah blah
Now if the jikes compilation skips the AM_LIBTOOL macro, the variables
are not resolved and the resulting Makefile includes this lovelly target:
:
blah blah
Of course running "make" on such a broken Makefile does not help in
compiling anything!
I assume, the ultimate solution would be for configure to skip AM_LIBTOOL
but initialise the missing Makefile.in variables with
reasonable defaults. The values are mostly irrelevent as these targets are
not executed anyway. The only requirement is a Makefile free of syntax
errors.
At the time I gave up! Considering that Ant is so simple (and fast) to use, I
decided it didn't worse the effort and time, and went on using Ant
scripts for bytecode compilation and specific GCJ Makefile.am for GCJ
native compilation.
Olivier
--
----------------------------------------------------------------------
Olivier Louchart-Fletcher
- [GNU Crypto] MD2 hash, Mark Wielaard, 2002/10/18
- Re: [GNU Crypto] MD2 hash, Casey Marshall, 2002/10/20
- Re: [GNU Crypto] MD2 hash, Raif S. Naffah, 2002/10/20
- Re: [GNU Crypto] MD2 hash, Casey Marshall, 2002/10/22
- Re: [GNU Crypto] MD2 hash, Raif S. Naffah, 2002/10/22
- Re: GCJ build (was Re: [GNU Crypto] MD2 hash), Olivier LF, 2002/10/22
- Re: GCJ build (was Re: [GNU Crypto] MD2 hash), Raif S. Naffah, 2002/10/23
- [GNU Crypto] Re: GCJ build,
Olivier LF <=
- Re: [GNU Crypto] Re: GCJ build, Raif S. Naffah, 2002/10/25
- Re: [GNU Crypto] Re: GCJ build, Raif S. Naffah, 2002/10/25
- Re: [GNU Crypto] Re: GCJ build, Olivier LF, 2002/10/26
- Re: [GNU Crypto] Re: GCJ build, Raif S. Naffah, 2002/10/27
- Re: [GNU Crypto] Re: GCJ build, Olivier LF, 2002/10/27
- Re: [GNU Crypto] Re: GCJ build, Raif S. Naffah, 2002/10/27
- Re: [GNU Crypto] MD2 hash, Casey Marshall, 2002/10/25
- Re: [GNU Crypto] MD2 hash, Raif S. Naffah, 2002/10/25