* Manuel A. Fernandez Montecelo <[email protected]>, 2012-02-05, 23:02:
So I should remove "Multi-Arch: same" line from the control and that's it,
Yes, that's the simplest way to fix it.
[...]
If you are feeling adventurous, you might want to keep M-A:same and make the header file identical across architectures, e.g. by exploiting the fact that GCC defines the __BYTE_ORDER__ macro.

Do you recommend this?

No be honest, no. :)

I don't like to diverge from upstream if there's no compelling reason to do so, especially if they are not quick to attend these kind of requests and so we have to carry patches for ages.

That's reasonable.

Also, is this macro supported by other compilers which can become more mainstream in the near future?

I don't know that.

What about pkg-config files? Is it only an issue if they have different content across architectures, or the mere fact of being shipped in a /usr/lib/TRIPLET/pkgconfig dir renders the package "non-Multi-Arch: same"?

pkg-config files are fine. The purpose of /usr/lib/TRIPLET/ is precisely to store architecture-specific files.

A while ago I announced existences of a tracker for problems like that[0] (which I'm use to file bugs when I'm bored :P).

So a way to generate this is, after building in the buildd machines, download all -dev packages and unpack them in different dirs for arches, then diff them?

This is more or less what my script do, yes. (In fact it takes a shortcut: rather than unpack packages, it only extract the md5sums control file and compares this. One can't detect all the problems[1] using this method, but in practice it's good enough.)

It's not connected to the PTS, though.

[0] http://lists.debian.org/debian-devel/2011/11/msg00508.html

So if it's reliable, it would be a good idea to add it to some of these tools :-)

Sure. Have you just volunteered to implement it? ;)


[1] E.g.: not all packages have md5sums; information about symlinks is missing.

--
Jakub Wilk



--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]

Reply via email to