Re: libtool woes

[Jukka Santala <jsantala at tml dot hut dot fi>]

On Fri, 1 Feb 2002, Alexandre Petit-Bianco wrote:
> Yes please. Don't include the ChangeLog entries as they often don't
> apply cleanly. Listing the ChangeLog entry and then the relevant hunks
> is the preferred form of patch submission (this is not the reason the
> patch wasn't applied in a timely fashion, BTW. We're all just really
> busy. Sorry about that.)

Yeah, I'm familiar with that convention, but consider it rather silly.
Especially with projects like rhug with literally dozens of ChangeLog's to
maintain, the duplicated effort both on the patch-author and and
patch-applier to deal with separate ChangeLog entries is quite a lot. And
if the patch to a ChangeLog fails to apply cleanly, you're basically left
with the separate entry to paste in anyway.

Ofcourse, the reason all the chunks failed, and not just the ChangeLog
entries, probably has more to do with the CVS diff format than any
ChangeLog changes itself. Some patch utilities apply them cleanly, others
are led to try to apply it to a wrong file if one with the same name
exists in the root directory where it's ran. And remember, this change
involves the Makefile.in file in every one of the 16 directories, each one
with their invidual ChangeLog entries ;) So it's getting to be quite a
drag to do this by diffs.

So anyway, here's the patch again, in plain patch format, which applies
cleanly with -p1 or from the directory with "rhug" tree in it. For what it
is worth, I separated the patches to all the 17 ChangeLog's into a
separate patch, but didn't feel like providing alternate text-copy of each
to go with the chunks they apply to. (Most directories have slightly
different ChangeLog entries or at least different timestamp format(!))

I added the fix for the AM_GCJFLAGS problem (Makefile's using the
user-flag GCJFLAGS instead, preventing the user from giving alternative
flags) as well to avoid the need for a separate patch to apply for that. I
also participated in little historical revisionism, since I didn't feel
like updating the ChangeLog entry timestamp in each affected directory,
since it's practically same anyway. The normal caveats apply; I did
rebuild the build-scripts, but since my tools are different version, that
would generate some ten megabytes of diffs, hence hopefully whoever
applies the patches will re-generate them ;)

 -Jukka Santala

Attachment: rhug-cross-changelog.patch
Description: Text document

Attachment: rhug-cross-build.patch
Description: Text document