-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 On 15/06/14 05:42 AM, Michał Górny wrote: > Dnia 2014-06-15, o godz. 16:06:57 "Vadim A. Misbakh-Soloviov" > <m...@mva.name> napisał(a): > >> My idea is to allow failing for some patches without breaking >> build at all. And, in parallel, to add groupping. >> >> How I imagine that: >> >> etc/portage/patches/app-cat/<name>/ | | - group_name/ | | | >> |- 01_foo.patch | |- 02_bar.patch | |- <...> | |- >> 01_moo.patch |- 99_meow.patch >> >> Where every first-level piece (patch or group) in >> ```etc/portage/patches/app-cat/<name>/``` MAY tolerably fail (not >> causing "die" for emerge), but if one of the patches inside the >> group fails, then group MUST NOT be applied at all (and all >> previously applied patches from this group MUST be reversed). > > Just don't. > > Or more specifically: it's not worth the effort, the extra > complexity, the confusion and the wholesale mess involved. >
Agreed. patches, or groups thereof, should not ever be fail'able without also stopping the emerge process. And the whole if-one-fails-then-revert-the-group thing would be hell to implement. Even if the patch fails because it's determined to have been "already applied", there's no guarantee that this check is accurate (ie what upstream applied is the same as your patch). Best to just fail, to let users know they need to clean up their patches. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.22 (GNU/Linux) iF4EAREIAAYFAlOfB5kACgkQ2ugaI38ACPDDRgEAujxxI9LLTs8Bj+nNgGgUcG15 XLNXD3vtpzbVmtE6MsgBAKAGO4Ysjwt07uVMlXWNqQz31QRUza24/lIOkVafnTDd =5G8J -----END PGP SIGNATURE-----
