Makefile files are missing DESTDIR on some spots

[stick]

Makefiles are missing DESTDIR in some places

[stick]

patch

[pygi]

Thank you for your contribution. Your patch will be applied.

[scdbackup]

Sorry, that won't work - for autotools reasons:

You have to teach Makefile.am and not Makefile.in. The latter is generated by the ./bootstrap command.

And somebody has to teach me about why and what before i can agree to changes in Makefile.am and/or configure.ac. Both are very obscure and it is easy to use them for shooting ones own foot.

So for now: what is the patch supposed to change in comparison to the old state ? What problem shall get fixed ?

I acknowledge that much of our build system is questionable and needs skilled review. But we must take care not to break what is working for now.

[pygi]

This is indeed for release tarballs. Regardless, we should investigate and modify the patch to our svn build system if it is indeed useful.

[stick]

The main problem is when creating symlinks for xorrisofs.

When you use build in chroot the symlink is created out of chrooted environment (missing DESTDIR).

I do not know how and if the other forgotten DESTDIRs are critical.

[scdbackup]

The main problem is when creating symlinks for xorrisofs.

Aha. Then i am the one to blame. The links are created by the install-exec-hook of Makefile.am which are quite outside the "auto" part of autotools.

I'll have to do experiments. Any help is welcome. The job would be to modify the hooks in libisoburn/Makefile.am and in libisoburn/xorriso/xorriso_makefile_am.txt .

[scdbackup]

I introduced the $(DESTDIR) to the install-exec-hook: of libisoburn-0.2.3 and xorriso-0.2.3. It seems to work were it used to work before. (At that occasion i see how ugly the .PHONY buildstamp is at install time. I'll revoke that feature for now.)

So the SVN libisoburn/trunk version now should avoid that mistake in the handmade install-exec-hook. Mario: could you produce an inofficial tarball libisoburn-0.2.3.tar.gz and make it available to stick ? I'll upload a xorriso-0.2.3.tar.gz tarball soon.

---

But this here looks ok to me.

        test -z "$(libdir)" || $(mkdir_p) "$(DESTDIR)$(libdir)"

The test -z does not check the existence of a directory but the content of a string variable. So adding "$(DESTDIR)" could interfere unduely with that test.

We should leave that as is and hope that others tested it to be harmless.

[stick]

Argh, my bad. Sorry ...

[stick]

And no need for the tarball from my side ...

[scdbackup]

You are the one who can reproduce the problem. It slipped through all our tests so far. (How does one get non-empty content into DESTDIR ?)

So please try the xorriso-0.2.3 tarball which is uploaded meanwhile and/or the libisoburn tarball when it becomes available. They are still very similar to 0.2.2 . No other significant development has happend since saturday.

If this all works well for you then i will probably patch the release tarballs to .pl01. So an OK by you would be very valuable.

[scdbackup]

I forgot to post the URL

http://scdbackup.sourceforge.net/xorriso-0.2.3.tar.gz

[pygi]

I will make the libisoburn tarball available later today, since it's past 1AM now :)

[scdbackup]

libisoburn-0.2.2.pl01 and xorriso-0.2.2.pl01 have been released and should be correct in that aspect. Case closed.