2008/1/24, Alessio Bolognino <themolok.ml@gmail.com>:
I'm who wrote the patch, so I think I should protect my baby a little bit, even if I am not an strong supporter of it:
On Wed 2008-01-23 17:45 , Dan McGee wrote:
[...]
I'm going to be the hardass here and first ask which patch we are considering above. The global variables thing worries me a bit.
Global variables work 90% of time, they don't work if you mess with them too much (and it's documented)
I feel like the advantages of having it in a second file, which is copied directly to the package, outweighs the benefits. Notice that with the above method, zero comments are copied over as well.
True, but usually people look at the source PKGBUILD, not at /var/lib/pacman/wherever/is/the/post-install
This could confuse people. I already mentioned the global variables thing as a drawback, in addition to using subfunctions (see the pacman.install file). It isn't that these couldn't be solved by simply sticking with an external install file, but the confusion of educating users then comes into play. "Hey, you can put install functions in your PKGBUILD, but ONLY if they don't use variables, you don't mind missing comments, you don't try to actually use subfunctions to clean up the code...".
Maybe I'm just being a stickler here, but it seems like what we have now works quite well, at the expense of needing one extra file. It also *clearly* seperates build-time operations from install-time ones, which I can see being quite confusing to first-time PKGBUILD writers.
I agree that an external file is more straightforward, but as you said, the patch only adds a feature, it's fully backward compatible.
I wrote this mail also to say that I'm not going to develop this patch anymore, so if it doesn't apply or makes your systems explode, don't ask me to fix it because I don't have time/will to do it.
Trying to resurrect this... Dan, can this be merged for 3.2? -- Roman Kyrylych (Роман Кирилич)