[pacman-dev] [PATCH] makepkg: allow epoch to be overridden

[Allan McRae]

On 23/07/11 00:13, Dave Reisner wrote:
> On Sat, Jul 23, 2011 at 12:10:32AM +1000, Allan McRae wrote:
>> On 22/07/11 23:55, Dave Reisner wrote:
>>> On Fri, Jul 22, 2011 at 10:04:31PM +1000, Allan McRae wrote:
>>>> We can override pkgver and pkgrel so it is only logical to add epoch
>>>> to that list
>>>>
>>>> Signed-off-by: Allan McRae<allan at archlinux.org>
>>>> ---
>>>>   scripts/makepkg.sh.in |   17 ++++++++++-------
>>>>   1 files changed, 10 insertions(+), 7 deletions(-)
>>>>
>>>> diff --git a/scripts/makepkg.sh.in b/scripts/makepkg.sh.in
>>>> index 4b8f167..88afe7b 100644
>>>> --- a/scripts/makepkg.sh.in
>>>> +++ b/scripts/makepkg.sh.in
>>>> @@ -44,9 +44,9 @@ startdir="$PWD"
>>>>
>>>>   packaging_options=('strip' 'docs' 'libtool' 'emptydirs' 'zipman' 'purge' 'upx')
>>>>   other_options=('ccache' 'distcc' 'buildflags' 'makeflags')
>>>> -splitpkg_overrides=('pkgver' 'pkgrel' 'pkgdesc' 'arch' 'license' 'groups' \
>>>> -                    'depends' 'optdepends' 'provides' 'conflicts' 'replaces' \
>>>> -                    'backup' 'options' 'install' 'changelog')
>>>> +splitpkg_overrides=('pkgver' 'pkgrel' 'epoch' 'pkgdesc' 'arch' 'license' \
>>>> +                    'groups' 'depends' 'optdepends' 'provides' 'conflicts' \
>>>> +                    'replaces' 'backup' 'options' 'install' 'changelog')
>>>>   readonly -a packaging_options other_options splitpkg_overrides
>>>>
>>>>   # Options
>>>> @@ -1455,10 +1455,13 @@ check_sanity() {
>>>>   		fi
>>>>   	done || ret=1
>>>>
>>>> -	if [[ ! $epoch =~ ^[0-9]*$ ]]; then
>>>> -		error "$(gettext "%s must be an integer.")" "epoch"
>>>> -		ret=1
>>>> -	fi
>>>> +	awk -F'=' '/^[[:space:]]*epoch=/ { $1=""; print $0 }' "$BUILDFILE" |
>>>> +	while read i; do
>>>> +		if [[ ! $i =~ ^[0-9]*$ ]]; then
>>>> +			error "$(gettext "%s must be an integer.")" "epoch"
>>>> +			return 1
>>>> +		fi
>>>> +	done || ret=1
>>>>
>>>>   	if [[ $arch != 'any' ]]; then
>>>>   		if ! in_array $CARCH ${arch[@]}; then
>>>> --
>>>> 1.7.6
>>>>
>>>>
>>>
>>> I see 2 issues:
>>> 1) s/return/exit/ or remove the '|| ret=1'. Both are not needed. return,
>>> even from inside a subshell will work as expected.
>>
>> Does either of those options work?  We want only ret=1 set on
>> failure and not exiting from the function so the rest of the checks
>> are performed.
>>
>
> Then you want to 'exit 1' from the subshell so that the '|| ret=1' is
> triggered.
>

As opposed to now when it is triggered?  Not saying that exit is not the 
better way to go, just that return works...

Allan