------- Comment #4 from rob1weld at aol dot com 2007-07-09 21:01 ------- >Comment #2 From Andrew Pinski 2007-07-09 02:53 [reply] ------- >No include for <string.h> or <cstring>.
Andrew, I'm not an expert at C++ but I did my best to attempt to make a couple of reduced testcases where I felt I was able. BOTH those testcases _do_ use "#include <string>". In addition both my testcases (and HPCToolkit) compile with versions of GCC < 4.3 . > All of these look like missing includes in the source No. (my testcases are not.) (see answer to Paolo below). > and nothing wrong with GCC 4.3.0. Maybe. Either GCC 4.3 is incorrect and all the prior versions are OK, _or_ the other way around. Between the people at Rice and you, you know more about C++ than me. >Comment #3 From Paolo Carlini 2007-07-09 08:43 [reply] ------- >Note that in 4.3 the header dependencies have been streamlined and it's well >possible that some projects around are failing to include required headers. >Please double check in this light, thanks. I'm building 4.2.1 (with concept checking off) so I can build HPCToolkit without any trouble. While that is running I'll open another window and go back and play with HPCToolkit some more (using 4.3). >dependencies have been streamlined That will break a lot of code. Asuming that 4.3 is adhearing to _all_ C++ standards correctly and that HPCToolkit (and many other programs) are poorly written it would be kind to have a "-fuse-old-headers" switch. Fixing versions of GCC < 4.3 to have correct header dependancies would only cause more programs to fail to compile. That's not 'our' fault but sometimes we need to be accomidating (Look at all the switches in gcc.info 'C++ Dialect Options' we have just to be compatible with old code - may need one more). There exists TWO issues not dealt with in your replies (see above): 1): "There is something about the way GCC 4.3 builds "gij" that does not let it run .jar files that work perfectly on the WinXP platform with Internet Explorer." 2): The ".ii" files that GCC produces are not 'all inclusive' - you can't move them and then simply compile them in a different directory (like ".i" files). How are we supposed to create an ".ii" file and attach it for an expert to study? (as is suggested we do for "C" errors and ".i" files). Maybe the streamlining has caused this. Paolo, I'm doing that now. Back before tommorrow. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32683
