[Xcb] [PATCH libxcb] config: default option for enable-dri3 is not implemented

[josh at joshtriplett.org]

On Tue, Mar 25, 2014 at 05:31:42PM -0400, Gaetan Nadon wrote:
> On 14-03-24 09:55 AM, Josh Triplett wrote:
> > Could you also provide a patch fixing the bug you mention in your commit
> > message, that an --enable-dri3 gets silently ignored if !sendfds?
> Sure. Here's the spec, to make sure I understand.
> 
>     User input:                sendfds state    Result:
>     -----------------------------------------------------------------------------
>       <nothing>                sendfds=yes      DRI3 is enabled
>       <nothing>                sendfds=no       DRI3 is disabled
>       --enable-dri3=auto       sendfds=yes      DRI3 is enabled
>       --enable-dri3=auto       sendfds=no       DRI3 is disabled
> 
>       --enable-dri3            sendfds=yes      DRI3 is enabled
>       --enable-dri3            sendfds=no       Error Msg Configuration
>     ABORTS
>       --enable-dri3=yes        sendfds=yes      DRI3 is enabled
>       --enable-dri3=yes        sendfds=no       Error Msg Configuration
>     ABORTS
> 
>       --enable-dri3=no         sendfds=yes      DRI3 is disabled
>       --enable-dri3=no         sendfds=no       DRI3 is disabled
>       --disable-dri3           sendfds=yes      DRI3 is disabled
>       --disable-dri3           sendfds=no       DRI3 is disabled
> 
>     User input "<nothing>" and "--enable-dri3=auto" have the same
>     automake semantic
>     User input "--enable-dri3" and "--enable-dri3=yes" have the same
>     automake semantic
>     User input "--enable-dri3=no" and "--disable-dri3" have the same
>     automake semantic

Correct.

>     Additional problem
>     -----------------
>     There is a flaw in the current code. If user input is "--enable-sendfds"
>     or "--enable-sendfds=yes", it is taken at face value and the
>     presence of the
>     required library is not checked.(See AC_SEARCH_LIBS line 98).
> 
>     Clarifying --enable-sendfds option
>     ----------------------------------
>     If sendmsg is only used for dri3, perhaps the --enable-sendfds
>     option is redundant.
>     Would a builder configure libxcb with --enable-dri3 and
>     --disable-sendfds?
>     Or --disable-dri3 and --enable-sendfds together?
> 
>     If sendfds is used for other purposes than dri3 (now or in the
>     future), then it is ok.

I don't see much point in having a separate option for sendfds, rather
than just always autodetecting it.

- Josh Triplett