CGAL users discussion list

[Sylvain Pion <>]

Pavlin, Andrew a écrit :
> Greetings.
>
> We had an unfortunate install problem with CGAL 3.3 when storing the 
> CGAL source files in a ClearCase VOB. Because we weren't modifying any 
> of the CGAL source files, we left them all not checked out (and 
> therefore with read-only permissions), and attempted to run the 
> install_cgal script to customize it to our development platform (we 
> currently use three platforms: Microsoft Windows XP with Cygwin and g++ 
> 3.3.3, Sun Solaris 10 with Sun Studio 11, and Red Hat Enterprise Linux 
> with gcc 3.4, with ClearCase installed on all the development 
> platforms). The install script started reporting errors about the time 
> it got down to testing BOOST, wth repeated errors:
>
> ./install_cgal:2955: Permission denied
>
> When the script completed, it claimed that BOOST could not be found, 
> despite the fact that I had passed it in on the command line, and (in 
> interactive mode) had specified it again before running the Test and 
> Setup phase.
>
> The problem turned out to be that all of the configuration testing 
> source files in CGAL/config/testfiles had permissions 444 (read-only) 
> because of being ClearCase controlled files. So the first test file was 
> successfully copied to tmp_file.cpp, and all the remaining ones failed 
> because the cp command couldn't overwrite the existing file (which 
> preserved the permissions of the first copied file, rather than being 
> owner-write access).
>
> I managed to get the install_cgal script to work in this environment by 
> checking it out, and inserting the line
>
> ${_chmod} u+w ${TMP_FILE_CXX}.cpp
>
> after every occurrence of a ${_cp} sometestfile.cpp ${TMP_FILE_CXX}.cpp. 

Another option would be to avoid the copy, and compile the original file 
directly.  I haven't looked at the code so maybe there are reasons why
it would be not easy.
But, it any case, we are in the process of replacing the install_cgal
script by CMake, so I think it's not worth working on this dead end now.


> Note that this file still has the problem with ignoring 
> --BOOST_INCL_DIR; I have not tried to fix that problem, since I am told 
> it is already a known bug.

Maybe the switch to CMake will fix this as well...
(and we'll have new fresh bugs to fix :-)

--
Sylvain Pion
INRIA Sophia-Antipolis
Geometrica Project-Team
CGAL, http://cgal.org/