c/8597: Enhancement request

[Steven Overhauser <spo@purple.int.infinisim.com>]

>Number:         8597
>Category:       c
>Synopsis:       gcov related files getting overwritten
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    unassigned
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Fri Nov 15 19:26:00 PST 2002
>Closed-Date:
>Last-Modified:
>Originator:     Steven Overhauser <spo@ngls.net>
>Release:        3.2
>Organization:
ngls.net
>Environment:
System: Linux purple 2.4.18 #2 Thu Mar 7 11:18:15 PST 2002 i686 unknown
Architecture: i686

	
host: i686-pc-linux-gnu
build: i686-pc-linux-gnu
target: i686-pc-linux-gnu
configured with: ../configure --prefix=/usr/local/dev/Linux_i686/gcc3.2 --enable-shared --enable-threads=posix --enable-version-specific-runtime-libs
>Description:

  The gcov related files (*.bb, *.bbg, and *.da) 
end up in the current directory where the compilation
actually took place.  This is a problem if you build
multiple architectures from the same spot.  (As the
make system I use does.)
  This also makes one very crowded directory when
compiling an entire hierarchy of files from the
root of the hierarchy.  Also, there would be a 
problem with multiple source files with the same
name in different directories.

>How-To-Repeat:

On a linux box:
gcc -fprofile-arcs -ftest-coverage -g path/file.cpp -o srcPath/linux/file.o
And on a solaris box in the same directory:
gcc -fprofile-arcs -ftest-coverage -g path/file.cpp -o srcPath/solaris/file.o

The solaris .bb and .bbg would have just overwritten the
linux based files.

>Fix:

  I believe it would make more sense to put these gcov
files into the same directory as the object file that
gets created.  This would fix both the multiple architecture
problem and the similarly named file problem mentioned
above.
  I have tried to hack the source code myself, but
haven't managed to get it to work yet.  I've only
been able to get the files to go into the "srcPath"
directory (as in the above example) and not the
object file's directory.
  By the way, my hack also included ignoring creating
.bb and .bbg files for files whose paths started
with "/usr" so that I did not get those files for
all of the system header files.  It seems to me that
this should some sort of option.
>Release-Note:
>Audit-Trail:
>Unformatted: