aboutsummaryrefslogtreecommitdiffstats
path: root/src/nativewindow/classes
diff options
context:
space:
mode:
authorSven Gothel <[email protected]>2023-01-16 01:02:47 +0100
committerSven Gothel <[email protected]>2023-01-16 01:02:47 +0100
commita8bb3295b467d35d26bc33c11797b516ba775903 (patch)
tree576e245bcb9585944fc14a328b4c4213b9658992 /src/nativewindow/classes
parenta1514cb44f2d9e82e737cc6a6f7e3477bb0fbbcc (diff)
X11GLXContext.getPlatformExtensionsStringImpl(): Add comment from pull request regarding GLX extensions
https://github.com/sgothel/jogl/pull/107 Return either glXQueryClientString or glXQueryExtensionsString when getting the GLX extensions ndjojo: glXQueryExtensionsString will make a request for glXQueryServerString if needed and will append the necessary client-side extensions. This doc, under the section "Using GLS Extensions", also suggests checking the glXQueryExtensionsString before using GLX extensions. +++ aschleck: For some more context this came up with the release of Mesa 20.3.0, which has client support for GLX_EXT_swap_control but no server support. The current JOGL behavior of appending the client extensions to the server extensions is incorrect. They should instead be intersected (with client-only extensions then appended) as the doc Nicole linked above says, which is precisely what glXQueryExtensionsString does. With the current extension querying behavior JOGL thinks glXSwapIntervalEXT is available under Mesa/llvmpipe even though it is not, causing a segfault at JOGL initialization time. I originally filed this as a Mesa bug (https://gitlab.freedesktop.org/mesa/mesa/-/issues/4128) along with some code that repro'd JOGL's checking behavior but it became apparent that Mesa is fine and the checking behavior is incorrect.
Diffstat (limited to 'src/nativewindow/classes')
0 files changed, 0 insertions, 0 deletions