DICOM @ OFFIS

Hi Michael,

The latest snapshot (dcmtk-3.5.5_20100504) still resets the Microsoft Visual C++ compiler flags, even when we specify different compiler flags within cmake-gui.

Did you still have time to discuss the CMakeLists, as you planned to do after the easter days?

When a memory allocation failure occurs during decoding, we sometimes get into a deadlock (*), during a later stage. For example, when we try to cleanup() a decoder registration. The deadlock is caused by the fact that the operator new might throw an exception (std::bad_alloc). If this happens just ...

Hi Jörg, FYI: We will disable this warning for the next snapshot Thanks! You wrote before: However, I'm not quite sure whether this kind of array initialization really works with all kind of compilers. FYI, I'm currently testing this kind of initialization ("value-initialization"), for var...

Hi Jörg, I can already tell you that my colleague Jeroen Eggermont was using the February snapshot, while I'm using the snapshot of March 2010 (dcmtk-3.5.5_20100326). We both encountered the very same crash. Is the explanation clear enough to you? Jeroen observed a truncation during the calculation ...

Thank you, Michael. Of course, we wouldn't mind if you would just remove all those SET commands: # settings for Microsoft Visual C IF(CMAKE_C_COMPILER MATCHES "CL|cl") IF(NOT WITH_SHARED_LIBS) IF(CMAKE_GENERATOR STREQUAL "Visual Studio 6") SET(CMAKE_C_FLAGS "/nologo /W3 /GX ...

We are very pleased about the snapshots being regularly updated. However, it also causes some overhead on our side, having to locally patch each snapshot before we can use it. For example, we use different Visual C++ 2008 compiler flags as we statically link to the Microsoft Runtime Library (/MD, /M...

warning C4351: new behavior: elements of array 'DiARGBImage::Palette' will be default initialized warning C4351: new behavior: elements of array 'DiPaletteImage::Palette' will be default initialized That's my favorite compiler warning ;-) Telling us that the compiler actually does the right thing, ...

Thanks, Jörg. I hadn't yet spotted the equivalent DiARGBImage bug. Value-initialization of an array, as I suggested by doing Palette(), should work fine on any modern C++ compiler. But of course, I wouldn't mind if you'd initialize the array elements in a different way, e.g., by doing Palette[0] = P...

In the latest snapshot (dcmtk-3.5.5_20100226) DiPaletteImage (dipalimg.h) contains an array of pointers, Palette: /// three color palettes for R, G and B DiLookupTable *Palette[3]; Unfortunately, the constructor of DiPaletteImage does not always initialize the elements of this array. Quoting from di...

Jörg Riesmeier wrote:

Will be fixed in the next snapshot.

Thanks!

dcmdata\include\dcmtk\dcmdata\dchashdi.h has the following function: OFBool operator!=(const DcmHashDictIterator& x) const { return (hindex != hindex) || (iter != x.iter); } This is clearly a bug, as (hindex != hindex) is entirely useless. Instead it should say (hindex != x.hindex) . The bug is ...

Michael Onken wrote:

I've send you an email how to download a recent CVS snapshot.

Thank you very much, Michael!

-- Niels

Michael Onken wrote:

fixed in CVS

Cool! Is the DCMTK CVS publicly accessible? Otherwise, can you please send me the fixed version, so that we can use it here at the LKEB in Leiden?

There's a mail address of mine at www.xs4all.nl/~nd/dekkerware

Jörg Riesmeier wrote: you should probably use loadFile() with read mode ERM_fileOnly. Thanks. I just tried ERM_fileOnly, but unfortunately I still got the assert failures. I guess that assertions only occur in debug mode, right? Right. But in release mode, DcmItem::foundVR might still crash, when th...

Thanks for your reply, Michael! Michael Onken (OFFIS DICOM Team) wrote: I would say it makes more sense to change the API of foundVR() from accepting char* to Uint8* instead. This would implicate corresponding changes in checkTransferSyntax() to read the characters from the input stream into an Uint...