Bug 685708 – sys-apps/portage:: emirrordist ignores USE-conditional RESTRICTs

• Product: Portage Development
• Status: CONFIRMED ---

[Description Sergei Trofimovich (RETIRED) 2019-05-12 10:40:41 UTC]

Noticed on two ebuilds when ran full emirrordist:

$ emirrordist --verbose --mirror --tries 1 --repo gentoo --distfiles $(portageq envvar DISTDIR)

[INFO] failure: games-action/descent1-data-1.4a setup_descent_1.4a_(16596).exe  no fetchable uris
[INFO] failure: games-action/descent2-data-1.2  setup_descent_2_1.1_(16596).exe no fetchable uris

AFAIU normally these would be fixed by RESTRICT="fetch mirror".

Ebuilds have the following:

$ cat descent1-data-1.4a.ebuild
  MY_EXE="setup_descent_1.4a_(16596).exe"
  SRC_URI="cdinstall? ( http://www.dxx-rebirth.com/download/dxx/misc/descent-game-content-10to14a-patch.zip )
        !cdinstall? ( ${MY_EXE} )"
  RESTRICT="bindist !cdinstall? ( fetch )"

$ cat descent2-data-1.2.ebuild

  MY_PATCH="http://www.dxx-rebirth.com/download/dxx/misc/d2xptch12.tgz"
  MY_EXE="setup_descent_2_1.1_(16596).exe"
  SRC_URI="cdinstall? ( ${MY_PATCH} )
          !cdinstall? ( ${MY_EXE} )"
  RESTRICT="bindist !cdinstall? ( fetch )"

Adding "mirror" into RESTRICT under USE-condition does not help.

Is it an ebuild bug or emirrordist bug?

[Comment 1 Zac Medico 2019-05-13 03:27:50 UTC]

The SRC_URI evaluation is oversimplified, it uses use_reduce with matchall=True like this:

> >>> use_reduce('cdinstall? ( http://www.dxx-rebirth.com/download/dxx/misc/descent-game-content-10to14a-patch.zip ) !cdinstall? ( setup_descent_1.4a_(16596).exe )', uselist=None, matchall=True, eapi='6')
> ['http://www.dxx-rebirth.com/download/dxx/misc/descent-game-content-10to14a-patch.zip', 'setup_descent_1.4a_(16596).exe']

We make it loop through all combinations of states for the USE flags found in RESTRICT, and evaluate SRC_URI only for those combinations where fetch is not restricted.

[Comment 2 Zac Medico 2019-05-17 23:18:25 UTC]

The number of USE combinations could be prohibitively large in some cases (see bug 374397 for example), so in those cases we'll fallback to the matchall behavior.