[pacman-dev] [GIT] The official pacman repository branch, master, updated. v3.5.3-548-g9a40927

[Dave Reisner]

On Tue, Aug 09, 2011 at 08:35:35AM +1000, Allan McRae wrote:
> On 09/08/11 08:18, Dan McGee wrote:
> >commit 1a919a11b85cb882bf5e632036a9030e4a98aba0
> >Author: Dave Reisner<dreisner at archlinux.org>
> >Date:   Mon Aug 8 17:20:53 2011 -0400
> >
> >     makepkg: ignore epoch when undeclared
> >
> >     In this case, we skip the epoch versioning entirely, as if it were
> >     declared as 0.
> >
> >     Prevents errors such as:
> >
> >     /usr/bin/makepkg: line 244: ((: !  : syntax error: operand expected
> >     (error token is " ")
> >     ==>  Finished making: cower-git :20110808-1 (Mon Aug  8 17:17:27 EDT
> >     2011)
> >
> >     Signed-off-by: Dave Reisner<dreisner at archlinux.org>
> >     Signed-off-by: Dan McGee<dan at archlinux.org>
> >
> 
> 
> How do you ever get that?
> 
> # set defaults if they weren't specified in buildfile
> pkgbase=${pkgbase:-${pkgname[0]}}
> epoch=${epoch:-0}
> 
> So epoch should always be set...   perhaps the real bug should be fixed.
> 
> Allan
> 

Well, we're clearly inconsistant about it. I show some calls to
get_full_version returning epoch as 0 and some returning epoch as a null
string. It leads to a pkgver that's could possibly be something like
":20110808". Excerpts from my debugging:

==> Entering fakeroot environment...

EPOCH=|0|
FAKEROOTKEY=|1562776010|
==> Starting package()...

....stuff happens....

==> Creating package...
  -> Generating .PKGINFO file...

EPOCH=||
FAKEROOTKEY=|1562776010|
/home/noclaf/src/c/pacman/scripts/makepkg: line 246: ((: !  : syntax
error: operand expected (error token is "!  ")
  -> Compressing package...

EPOCH=||
FAKEROOTKEY=|1562776010|
/home/noclaf/src/c/pacman/scripts/makepkg: line 246: ((: !  : syntax
error: operand expected (error token is "!  ")
==> Leaving fakeroot environment.

EPOCH=|0|
FAKEROOTKEY=||
==> Finished making: cower-git 20110808-1 (Mon Aug  8 22:40:36 EDT 2011)


So is it fakeroot fucking with us here? Doesn't seem to be, but
somewhere along the line our epoch value is wrenched out from under us.

NB: (( ! $epoch )) isn't the same as (( ! epoch )). One expands to 0
when its unset and the other throws a syntax error.

d