[Mesa-dev] [PATCH] egl/dri: link directly to libglapi.so

[Tomasz Figa]

On Tue, Sep 26, 2017 at 6:44 PM, Eric Engestrom <eric at engestrom.ch> wrote:
> On 25 September 2017 12:01:33 BST, Tomasz Figa <tfiga at chromium.org> wrote:
>> On Mon, Sep 25, 2017 at 7:21 PM, Eric Engestrom
>> <eric.engestrom at imgtec.com> wrote:
>> > On Saturday, 2017-09-23 13:37:29 +0900, Tomasz Figa wrote:
>> >> On Wed, Sep 20, 2017 at 7:47 PM, Eric Engestrom
>> >> <eric.engestrom at imgtec.com> wrote:
>> >> > On Tuesday, 2017-09-19 17:19:59 +0000, Emil Velikov wrote:
>> >> >> From: Emil Velikov <emil.velikov at collabora.com>
>> >> >>
>> >> >> In order to build EGL, one has to use shared glapi -
>> libglapi.so.
>> >> >>
>> >> >> Thus the dlopen/dlsym dance is no longer needed and we can link
>> to the
>> >> >> library directly.
>> >> >>
>> >> >> This allows us to remove a handful of platform specific names of
>> the
>> >> >> library.
>> >> >>
>> >> >> Cc: Jonathan Gray <jsg at jsg.id.au>
>> >> >> Cc: Jon Turney <jon.turney at dronecode.org.uk>
>> >> >> Cc: Julien Isorce <julien.isorce at gmail.com>
>> >> >> Cc: Rob Herring <robh at kernel.org>
>> >> >> Cc: Tomasz Figa <tfiga at chromium.org>
>> >> >> Signed-off-by: Emil Velikov <emil.velikov at collabora.com>
>> >> >
>> >> > Nice cleanup!
>> >> > Assuming the build systems stuff works (with Rob's suggested
>> change?):
>> >> > Reviewed-by: Eric Engestrom <eric.engestrom at imgtec.com>
>> >> >
>> >> > struct dri2_egl_driver now only contains the glFlush() pointer
>> and
>> >> > _EGLDriver.  Could we move the _glapi_get_proc_address() call to
>> >> > the two places that use glFlush() (ie. dri2_make_current() and
>> >> > dri2_client_wait_sync()), and get rid of this struct, as well as
>> >> > the whole dri2_load() function?
>> >> >
>> >> > I'm happy to do this, just want to check that it would be ok :)
>> >>
>> >> Wouldn't that mean going through the series of strcmp()s over all
>> the
>> >> symbols every time EGL wants to call glFlush?
>> >
>> > Well, I was thinking of making that call static, maybe even global
>> > (shared between the functions), so it would happen at most once for
>> each
>> > of the two functions, but yeah you're right, it would do it more
>> often
>> > than right now. Do you think this is an issue?
>> > I guess I'm dropping this idea if that's the case :]
>>
>> I guess a global function (let's say dri2_gl_flush()) that looks up
>> the symbol only at the first call and then calls it, wouldn't be so
>> bad.
>
> Gave this a stab last night, it would look like this:
> https://github.com/mesa3d/mesa/compare/master...1ace:wip/egl-drv-flatten
>
> (This doesn't compile right now; I'll rebase it when Emil's patch lands,
> but you can already see what I was thinking about.)

Looks quite good, thanks. I like the diffstat. ;)

Best regards,
Tomasz