[PATCH] QT 4.4.0-snapshot-20070420 [missing q_atomic_*]

[Thiago Macieira]

Stefan Teleman said:
> Hi.
>
> A few patches for QT 4.4.0 snapshot from 2007/04/20 are here:

Thank you, but please submit them to qt-bugs at trolltech.com, indicating
where they are useful (i.e., where they are missing right now).

> http://www.stefanteleman.org/qt4/
>
> There's a test case for the q_atomic* asm functions in there
> (qatomictestcase.c) (this is for the missing asm functions in
> src/corelib/arch).
>
> The assembler functions were tested on Solaris 10 SPARC 32-bit / X86
> 32- and 64- bit with Sun Studio 11 and GCC 3.4.3
> with /usr/ccs/bin/as. The SPARC assembler was tested for v8plus,
> v8plusa and v8plusb.

I have some reservations about the Intel ones, but I'll let someone more
experienced in those than I speak. They don't look atomic to me.

However, the use of quads in x86-64 seems wrong. The pointer being passed
in q_atomic_fetch_and_add_int is a 'volatile int *', so it's 32-bit, not
64-bit.

You look like you wrote q_atomic_fetch_and_add_long.

> There are two small patches for tools/qdbus/tools/qdbus and
> tools/qdbus/tools/qdbusxml2cpp.

Those look Ok to me, but they're only stylistic. Both tools have CONFIG +=
qdbus, which adds $QTDIR/include/QtDBus to the include path. So #include
<QtDBus> is acceptable, if slightly different from the other includes.

> I'll have another patch for qplatformdefs.h for Solaris tomorrow
> (QT_SOCKLEN_T type isn't set right on Solaris 10 and it causes
> compile failures with Studio -- unfortunately the correct type for
> QT_SOCKLEN_T can no longer be determined by _XOPEN_SOURCE >= 500 or
> _POSIX_PTHREAD_SEMANTICS being defined).

socklen_t is one of the major sources of incompatibilities in socket code...

-- 
  Thiago Macieira  -  thiago (AT) macieira.info - thiago (AT) kde.org
    PGP/GPG: 0x6EF45358; fingerprint:
    E067 918B B660 DBD1 105C  966C 33F5 F005 6EF4 5358