Created attachment 166203 [details]
small example to demonstrate the issue.

Each \cond command normally needs a closing \endcond. In my opinion the 
end of a file should automatically close all conditional blocks in case 
someone forgot the \endcond. Or in other words the scope of a 
\cond+\endcond block shall not span outside a file. Currently such block 
seems to continue also over a file end, which is useless in my opinion 
because you cannot control the order doxygen parses the files.

Here is some background:
A colleague forgot the \endcond in two or three files. This caused some 
confusion: Documentation of some items in some other files was missing. 
It took me some time to find out that the missing \endcond commands were 
the reason. Those conditional were false, so doxygen ignored everything 
from the \cond on. But because \endcond was missing doxygen ignored also 
other things in other files which I do not want to be ignored.

P.S.
I know that doxygen warns if \endcond is missing. But it is easy to 
oversee such warning.

P.P.S.
Dimitri confirmes that there sould be a protection in convertCppComments in commentcnv.l. But I found out that it does not work all the time. The bug seems to depend on the order of input file handling. I will attached a small example to demonstrate the issue. Please watch for the order of input files:

	# This input line triggers the defect:
	INPUT = cond.c testdef.c
	
	# This INPUT line does not trigger the defect and 
	# everything looks like it should:
	#INPUT = testdef.c cond.c