I noticed (using ldd, incidentally while testing the new mesa_dri_drivers.so) the classic drivers (radeon, r200, i915, i965, nouveau_vieux) links to all three of libdrm_intel / libdrm_nouveau / libdrm_radeon, while only the matching one should be needed. Gallium drivers (i915, r300, r600, radeonsi, nouveau) are OK, linking only to their libdrm.
It looks every classic driver (including swrast) now includes all the classic drivers, they are more or less a copy of mesa_dri_drivers.so.
(In reply to comment #1) > It looks every classic driver (including swrast) now includes all the > classic drivers, they are more or less a copy of mesa_dri_drivers.so. They are in fact exact copies -- they're hardlinks. The point of mega drivers was to link all of the (classic) drivers into a single file.
Thanks but it doesn't look they are hardlinks anyway, the md5sum all differ, also their size is slightly different.
(In reply to comment #2) > The point of mega drivers was to link all of the (classic) drivers into a > single file. ... to waste memory on runtime if you make use of packages provided by distributions which contain all classic drivers ... I bet the solution to make only required symbols PUBLIC in former libdricore and libgallium isn't that much worse like you propagate. I just want to mention that there was never a comparision to my patchset which closes a lot of symbols for libdricore's replacement: http://lists.freedesktop.org/archives/mesa-dev/2013-September/044593.html
(In reply to comment #3) > Thanks but it doesn't look they are hardlinks anyway, the md5sum all differ, > also their size is slightly different. Strange, that's not what I see on my system: mattst88@work-Thinkpad mesa % md5sum $(find -name '*_dri.so') e918941fd19d4afaeb5834e90c5f88a4 ./lib/swrast_dri.so e918941fd19d4afaeb5834e90c5f88a4 ./lib/r200_dri.so e918941fd19d4afaeb5834e90c5f88a4 ./lib/i915_dri.so e918941fd19d4afaeb5834e90c5f88a4 ./lib/i965_dri.so e918941fd19d4afaeb5834e90c5f88a4 ./lib/radeon_dri.so e918941fd19d4afaeb5834e90c5f88a4 ./install/usr/local/lib/dri/swrast_dri.so e918941fd19d4afaeb5834e90c5f88a4 ./install/usr/local/lib/dri/r200_dri.so e918941fd19d4afaeb5834e90c5f88a4 ./install/usr/local/lib/dri/i915_dri.so e918941fd19d4afaeb5834e90c5f88a4 ./install/usr/local/lib/dri/i965_dri.so e918941fd19d4afaeb5834e90c5f88a4 ./install/usr/local/lib/dri/radeon_dri.so
(In reply to comment #5) > (In reply to comment #3) > > Thanks but it doesn't look they are hardlinks anyway, the md5sum all differ, > > also their size is slightly different. > > Strange, that's not what I see on my system: 1) If someone cares about why my .so that differ my compiled drivers are here: https://launchpad.net/~oibaf/+archive/graphics-drivers/+files/libgl1-mesa-dri_10.0%7Egit1310251530.980755%2Bglvdpau%7Egd%7Er_i386.deb build log is here: https://launchpadlibrarian.net/154982264/buildlog_ubuntu-raring-i386.mesa_10.0~git1310251530.980755%2Bglvdpau~gd~r_UPLOADING.txt.gz 2) I would have expected that classic drivers would be built including only a single driver and only mesa_dri_driver.so would include all drivers. 3) If there is no chance for 2) why the build system don't create a symbolic link for classic drivers pointing to mesa_dri_driver.so rather than hard links (or similar copies like in my case)? 4) There should be a bit of documentation for all this.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.