diff options
| author | Christian Strømme <christian.stromme@digia.com> | 2012-11-23 12:29:36 (GMT) |
|---|---|---|
| committer | The Qt Project <gerrit-noreply@qt-project.org> | 2012-11-28 12:29:35 (GMT) |
| commit | e2f600de9553168b3eb4fcbd1d4c523a91ee33ef (patch) | |
| tree | 3752f7c44f436ba1808615e15ca5030022a0610d /src/corelib | |
| parent | c6e3cea94c2b36aebb76b90b8354a04caa2f9c60 (diff) | |
| download | Qt-e2f600de9553168b3eb4fcbd1d4c523a91ee33ef.zip Qt-e2f600de9553168b3eb4fcbd1d4c523a91ee33ef.tar.gz Qt-e2f600de9553168b3eb4fcbd1d4c523a91ee33ef.tar.bz2 | |
Fixes problem with single precision floats in QDataStream (Windows).
When the floating-point behavior in MSVC is set to "precise" (default),
assigning nan numbers to a float causes the bit pattern to be altered
(only affects 32bit builds). We should therefore not assign the swapped
value back to a float and use it.
Backported from Qt 5 (3f936e9094f3a6e4d76791c1eff7ae92f91b61ae)
Task-number: QTBUG-25950
Change-Id: I39725557ab89c978eb1c59a4c79df8a27ed70ecf
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
Diffstat (limited to 'src/corelib')
| -rw-r--r-- | src/corelib/io/qdatastream.cpp | 6 |
1 files changed, 5 insertions, 1 deletions
diff --git a/src/corelib/io/qdatastream.cpp b/src/corelib/io/qdatastream.cpp index 2e5a406..8b23479 100644 --- a/src/corelib/io/qdatastream.cpp +++ b/src/corelib/io/qdatastream.cpp @@ -1132,8 +1132,12 @@ QDataStream &QDataStream::operator<<(float f) } x; x.val1 = g; x.val2 = qbswap(x.val2); - g = x.val1; + + if (dev->write((char *)&x.val2, sizeof(float)) != sizeof(float)) + q_status = WriteFailed; + return *this; } + if (dev->write((char *)&g, sizeof(float)) != sizeof(float)) q_status = WriteFailed; return *this; |