bug report

[Thomas Desodt]

Hello !

this email to report a very strange bug, using g++ from the gcc-2.8.1 version on
Solaris.

I tried to compile one of my files (dcmodules.cc), and here is the compilation
error i got:

--- (compilation output start) ---

(cd include; make ARCH="" all)
echo Nothing to do for all
Nothing to do for all
(cd libsrc; make ARCH="" all)
c++     -DHAVE_CONFIG_H  -DNDEBUG  -c -I. -I. -I../include
-I../../config/include -I../../ofstd/include -I../../dcmdata/include
-I/usr/local/dicom/include/ofstd -I/usr/local/dicom/include/dcmdata
-I/export/home/micasX/development//src/include
-I/export/home/micasX/development//src/include/../packages/gif_util/fbm  \
         -O -D_REENTRANT -Wall -pedantic dcmodules.cc
dcmodules.cc: In function `static enum E_Condition
DcmModules::Add_NMMultiFrame_Module(class DcmDataset *, struct DataStore *)':
dcmodules.cc:2862: Internal compiler error.
dcmodules.cc:2862: Please submit a full bug report to address@hidden'.
*** Error code 1
make: Fatal error: Command failed for target `dcmodules.o'
Current working directory
/export/home/micasX/development/src/packages/dcmtk/dcmobjects/libsrc
*** Error code 1
make: Fatal error: Command failed for target `libsrc-all'

--- (compilation output stop) ---

(Note : i have included the file compilation_output.txt to give you a more
readable version of this output, and my dcmodules.cc file)

Now, the line 2862 says :
"numimg << numImages;",

where numimg is of type OFSTring (our redefinition of the 'string' class, and in
which i have overloaded the operator<< to append to the variable a string
version of the given numerical variable ), and numImages is of type UInt16
(redeclaration of unsigned short). I first thought that there was a problem in
my overloading of the operator<< function, BUT the line 2846 does not cause any
problem at all, and all the variables involved have exactly the same type. I
later tried another trick, which is to declare another variable called
numImages2 (line 2571), to which i assign the value of numImages (line 2772). If
on line 2862 i replace numImages by numImages2, then there is no compilation
error, and the program works fine. Likewise, if i replace numImages by
numFrames, it all works fine (except that the numerical value is not the right
one, of course !!!!!!).

I am very puzzled indeed.

So at the moment, using numImages2, my program works as it should, but i thought
that you might be interested in that strange bug... Or maybe you already know
it, and the correction exists !

Best regards,

Thomas.



NOTE : i have also tried to install version 3.1 and 3.1.1 of the gcc 'package',
but after 1.5 week of trying, using already made packages, getting the source
code and doing the biulding/compilation/installation myself, and always being
unable to get it to work, i finally gave up and reverted to my gcc-2.8.1 version
which (until 2 weeks ago)  always gave me full satisfaction !



--

Thomas Desodt [Software Engineer]
**************************************************
Park Medical Systems (UK) Ltd
phone : +44 (0)1252 37 67 37

   work email: address@hidden
private email: address@hidden
--

compilation_output.txt
Description: Text Data

dcmodules.cc
Description: Text document