| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
| |
Reviewed-by: Maurice
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Turning on LTCG affected too many projects, where customers applications
would take a long time linking, severly affecting their development time
(even though it was only added for release builds)
We turn it off by default, and add a -ltcg configuration option, and the
possibility to also do CONFIG+=ltcg in projects, should they not want it
for Qt, but in their own projects. (Same, they can build Qt with it, and
do CONFIG-=ltcg for their project)
Reviewed-by: andy
|
| |
|
|
|
|
|
|
| |
Better alternative to 9ffcf3737901c49acb1a1bc6b65572d195d78a5d
Task-number: 253052
Reviewed-by: Norwegian Rock Cat
Reviewed-by: MortenS
|
| |
|
|
|
|
| |
Forgot to override the linker, causing it to use gcc 4.0.
Reviewed-by: Richard Moe Gustavsen
|
| |
|
|
| |
Reviewed-by: mariusSO
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Commits 6e5774d84e7e3b68736f95fae09a084bd5b9ac7a and
014af3e99616f9ee13ca365566828f7daf77a7f6 broke the Qt build for Visual
Studio / IncrediBuild. The problem is as follows: QtWebKit has too many
entries in INCLUDE_PATHS. In moc.prf is a rule that builds a
mocinclude.tmp that contains command line options for moc.
moc gets then called with @mocinclude.tmp.
Above commits create an QMAKE_EXTRA_TARGET for the mocinclude.tmp,
which is necessary for parallel MinGW builds.
Alas, qmake -tp vc doesn't support QMAKE_EXTRA_TARGET.
See task 189914. Once this task is solved, this commit must be reverted.
Reviewed-by: joao
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
passing any `-j' to make.
Change 6e5774d84e7e3b68736f95fae09a084bd5b9ac7a made the problem much
less likely to occur, but in a debug_and_release build, debug and
release would share the same mocinclude.tmp. Therefore, if
exceptionally unlucky, the bug could still arise.
Put mocinclude.tmp under MOC_DIR so each exclusive build has its own.
Reviewed-by: Lincoln Ramsay
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
in parallel.
On Windows, when a project has a lot of INCLUDEPATHs, some extra logic
takes place to work around an issue of large command lines.
Instead of passing all the paths directly to moc.exe, the paths are
written to mocinclude.tmp which is then read from by moc.exe.
Prior to this change, every moc rule tries to write to mocinclude.tmp.
When running make with -j, this will happen in parallel, causing this
error message:
"The process cannot access the file because it is being used by another
process"
Change the logic so mocinclude.tmp is generated by its own rule.
Reviewed-by: Lincoln Ramsay
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
append the source dir to the include path, but only after the dirs with
the generated files.
this seems to have worked before only accidentally: the unqualified
default dirs were expanded to the source dir instead of the build dir,
but the build dir is added implicitly by default, so things magically
worked. now that we qualify the moc/ui dirs, projects relying on the
strange side effect suddenly break.
we should probably add the source dir to the include path by default,
but this coupling to uic/moc is closer to the historical behavior and
thus should be safer.
Reviewed-by: mariusSO
|
| |
|
|
| |
sometimes, even two reviews are not enough ...
|
| |
|
|
|
|
|
| |
this time, it can deal with directories which are absolute to start
with.
Reviewed-by: mariusSO
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
This reverts commit 7d2c8eb99c563b4fb236fe538123255f52f293a2.
This commit was not reviewed and breaks compile of QtNetwork on win32
with this error...
src\network\kernel\qauthenticator.cpp(55) : fatal error C1083: Cannot
open include file: '../3rdparty/des/des.cpp': No such file or
directory
... because the generated Makefile no longer contains src/network
as an include path.
|
| |
|
|
| |
This reverts commit 3aff9113a9702ea6f7e099a73136a718ae1b992f.
|
| |
|
|
|
|
|
|
|
| |
uic is set in default_pre. this leads to useless -I compiler flags when
no ui files are used in fact, so it would be nice to get rid of it in
cases where it is not used. for backwards compat, don't remove it from
the _pre, but add some magic to _post to remove it again if QT does not
contain gui. of course, we need a force_uic CONFIG to enable an
exception for QtGui itself ...
|
| | |
|
| |
|
