Use SDL for replay.

[armyofzin]

This change uses the SDL to create a window and handle events for the trace replayer (command line and editor). This change has a pervasive impact so any code reviews would be appreciated.

[mikesart]

I'm pretty sure we don't care about X11 anything once we get the SDL paths working. That's our current line of thinking at least...

[lawlove]

OK good - just thought I'd check.

On Wed, Jun 18, 2014 at 2:02 PM, Michael Sartain notifications@github.com
wrote:

I'm pretty sure we don't care about X11 anything once we get the SDL paths
working. That's our current line of thinking at least...

—
Reply to this email directly or view it on GitHub
#88 (comment).

[armyofzin]

Lawrence is right. There's a couple spots where I can just remove the X11 code instead of ifdef'ing it out. Working on that now. Should have it done in a few minutes.

The only caveat is the glxUseXfont stuff that I'm not sure how to work around with the SDL. I need to look at that more carefully. I left it as a TODO for now... hopefully that doesn't ruin anybody's day.

[mikesart]

If it's something not in SDL that makes sense to get into SDL, then we should try to do an official SDL patch. Checking with Sam / Ryan makes the most sense on those.

Otherwise we'll have to use SDL_GetWindowWMInfo() and have some code in there to handle the different platforms - something like I did in glxspheres.c. Maybe we should get this type of thing into vogl_port though?

[armyofzin]

Talk directly to the device through the SDL?

[mikesart]

Talk directly to the device through the SDL?

Yea - I'm talking about implementing vogl_replay_window::is_direct() and the font cache stuff by either getting a patch into SDL or putting the functionality into our vogl_port using SDL_GetWIndowWMInfo() and the native handles. Option #2 probably makes the most sense right now?

[armyofzin]

Alright. Do you want me to get that in this branch? Also I removed more X11/GLX code that Lawrence noticed.

[mikesart]

Do you want me to get that in this branch?

Would be nice to have a TODO comment in the is_direct() call, but otherwise no - doing it in another checkin is fine. Thanks Jeremy!

[mikesart]

I'm also getting a ton of warnings on Linux when building with Clang 3.4.1 (in the vogl chroot). Are you seeing this as well?

[armyofzin]

If you're referring to all the SDL-Doxygen warnings, then yeah I see those too. I think it might be something special clang is doing. It's on my todo list to silence those. It's coming directly from the SDL source.

[armyofzin]

Turns out Clang can parse Doxygen-style comments with the -Wdocumentation flag, which is damn cool. Using -Wno-documentation turns it off, but I don't think we want to do that? Ideally we just get this fixed in SDL.

[vMcJohn]

Adding a line-attached comment for testing purposes.