On 11/11/16 17:47, Giancarlo Razzolini via aur-general wrote:
Using a helper to confirm if a PKGBUILD is "ok" is so bad of a practice that I don't even know where to begin. Use the tools for the job, makepkg and, devtools. Let the helpers to deal with it. It's their job to make sure your package installs (provided you follow best practices).
I'm pretty sure I'd get in trouble if I flooded the git commit log trying to fix my mistakes through yaourt. That's like trying to fix a bug in a production environment... Fact is, I should have been more careful when building with makepkg ; yaourt was just intended as a final "polish" to make sure the "quick and easy" install worked. I should have noticed the error when building:
Packages re-use the building directory. This is were building in a clean chroot comes in handy. At minimum, use makepkg's -C option.
Yup, I'll make sure I confirm the build's validity in a clean(er) environment in the future. I did encounter the conflict earlier but I disregarded it because I thought it was simply due to the mess of a previous install. Using yaourt simply pointed it out again (probably because it is more systematic than I am), but at this point in the updating process, I agree that no errors should have occurred: On 11/11/16 18:02, Eli Schwartz via aur-general wrote:
yaourt can only tell you if you successfully pushed the update to the AUR website.
Exactly, that's what I intended to use it for. Instead, it pointed out a mistake I had missed earlier.
I strongly encourage you to learn how the package management and dependency/conflict resolution system works, it is extremely valuable information for any AUR user (whether they maintain packages or just download them).
I have a few things to review before I start working on actually significant package updates. I'll make sure I have a cleaner "AUR maintaining" environment to avoid missing such obvious errors in the future. I'll also make myself more acquainted with the whole process, since a better understanding of it would have signalled the error to me in time. Regarding boost-compute, I suppose I'll just file a deletion request for the package. Anyone using it has had to download boost anyway, and if they kept their system up-to-date, they should already have the compute code from the boost package in extra. Thanks again! -- Julien JPK (0xC3075A58) julienjpk@email.com