[buildd-tools-devel] schroot filesystem union support

[Roger Leigh]

On Sun, Mar 29, 2009 at 09:00:50PM +0200, Jan-Marek Glogowski wrote:
> On Sat, 28 Mar 2009, Roger Leigh wrote:
> 
> > [PATCH 05/22] Use 'make -C' and 'mkdir -p' in debian/rules
> >
> > mkdir -p not necessary since debian dir always exists
> 
> The "problem" is not the parent directory, but mkdir complains and fails,
> if the directory already exists.

Ah, right.  That makes sense.  I've committed this:

[ -d debian/build ] || mkdir debian/build

so it only does the mkdir if debian/build does not exist.

> I'm building my packages using "fakeroot debian/rules binary, which always
> "complains" about the existing debian/build directory.
> 
> > [PATCH 06/22] Adapt boostrap from libtoolize suggestions.
> >
> > ACLOCAL_AMFLAGS OK.
> > AC_CONFIG_MACRO_DIR not present, but does the same job as ACLOCAL_AMFLAGS,
> > is undocumented and should probably not be used (deprecated?).
> > autom4te.cache removal is unnecessary?  Seems fast enough without, so
> > should be OK.
> 
> I'm using libtoolize 2.2.4.  All changes are suggested by libtoolize.

OK.  While I'm happy with the ACLOCAL_AMFLAGS suggestion, I'm not
too keen on the AC_CONFIG_MACRO_DIR due to it not being present
in the autoconf manual.  In autoconf itself it's an empty macro;
I think ACLOCAL_AMFLAGS replaced it, and it's fine without it, so
I'd prefer to not add that.

It's quite possible that libtoolise is out of date WRT the
suggestions; it has to work with potentially much older versions of
the autotools, and we're using the latest stable releases of
everything.

> > [PATCH 08/22] Add another error detail and add log info functions.
> >
> > log_exception_info and log_ctty_exception_info are valid additions
> > to the API.  However, they were not added previously because
> > exceptions are by definition exceptional events which should
> > be at a minimum logged as a warning.  Reason information is
> > logged as an informational message.
> >
> > ==> log_exception_info and log_ctty_exception_info not committed
> 
> I'm using these in the next patch (don't fail on missing schroot.conf) to
> inform the user in verbose mode. As I get an exception from config.add
> when the file is missing but it should neither be an error nor a warning,
> I thought it's the easiest way to log it.

It's fine to log it this way.  I'm just wondering if it shouldn't be
classed as a warning rather than information, since the file is
missing (even if not strictly required).  This would fit into the
existing way we do things.


Regards,
Roger

-- 
  .''`.  Roger Leigh
 : :' :  Debian GNU/Linux             http://people.debian.org/~rleigh/
 `. `'   Printing on GNU/Linux?       http://gutenprint.sourceforge.net/
   `-    GPG Public Key: 0x25BFB848   Please GPG sign your mail.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: Digital signature
Url : http://lists.alioth.debian.org/pipermail/buildd-tools-devel/attachments/20090329/c1041dc8/attachment-0001.pgp