2014-10-05 13:12:29 by Havard Eidnes | Files touched by this commit (2) |
Log message: Add a patch segment to force the use of the IEEE 754 variant of the math library on NetBSD. |
2014-09-29 17:25:22 by Havard Eidnes | Files touched by this commit (3) |
Log message: Add a patch so that backtrace works when we hit an error: Adapt to NetBSD's backtrace_symbols() behaviour, using <> instead of (). Also increase symbol buffer -- some mangled symbols are Long. Also, temporarily switch to ia32.debug for the build and testing, to be replaced by native.* eventually. |
2014-09-24 09:58:30 by Havard Eidnes | Files touched by this commit (3) |
Log message: Upgrade to version 3.29.9.1. Fix the test target to be repeatable without errors... Add a separate test target -- "failing-tests", for tests which have failed on NetBSD for easier debugging. |
2014-08-31 21:55:49 by Havard Eidnes | Files touched by this commit (5) |
Log message: A few chanes rolled into one: * Use GYP_DEFINES instead of patching the .gyp file to use the pkgsrc-provided icu package * Disable the verbose build, but leave it commented out * On NetBSD/i386, build with -mstackrealign -mpreferred-stack-boundary=4 * Modify how the tests are invoked, and adapt after a "native" build * Use the newly created chromium-depot_tools package to make the tests run * Fix the platform/semaphore.cc fix for NetBSD > 6.99.4, include \ <sys/param.h> * Unlimit the data size to make more tests succeed PKGREVISION bumped. With this, out of 6652 tests expected to pass, 24 fail and 1 crash, down from initially 33 / 1. |
2014-08-27 11:12:05 by Havard Eidnes | Files touched by this commit (3) |
Log message: Fix the patch for missing sem_timedwait() by adding constification. |
2014-08-26 12:37:49 by Havard Eidnes | Files touched by this commit (2) |
Log message: Add a patch with a workaround for missing sem_timedwait() for NetBSD < 6.99.4, copied from the openjdk7 package. PKGREVISION not bumped, since this wouldn't build on earlier releases before, and for later versions this code is #if'ed out, so no change. |
2014-08-25 15:22:01 by Filip Hajny | Files touched by this commit (9) | |
Log message: Split v8 and v8-316, as they are too different now. Fix SunOS build for both after the last breaking update. Clean up Makefiles. |
2014-08-23 19:44:12 by Havard Eidnes | Files touched by this commit (8) | |
Log message: Update this so that I at least get a "d8" executable which can be started :) Local changes: * change MASTER_SITES to what appears to be in active use these days * update to version 3.28.62, the latest when I started on this update * Since devel/gyp has been updated so that if you include it's tool.mk, you get a "python" in the path, that substitution in Makefile has been removed * Instead of editing Makefile with sed, it's now patched; PYTONPATH is now also altered. * Use the pkgsrc-provided gyp and not the bundled version * Install a few more headers which come with the new version * Use "(intptr_t)pthread_self()" instead of syscall(SYS_gettid), the latter appears to be Linux-specific * Fix a type error by adding a cast to (void*) * Provide an ugly band-aid for a possibly missing PTHREAD_STACK_MIN constant (as on NetBSD) * Build agsinst the pkgsrc-provided icu, not the bundled third-party version, to avoid file name and/or version clashes * Do REPLACE_PYTHON on the various *.py scripts * Remove the workaround for thin archives, that's now done by the pkgsrc-provided gyp Not sure the rpath editing is needed or correct; we'll see. |