Manoj Srivastava <[EMAIL PROTECTED]> wrote: > Yes, the new flex is no longer lex/POSIX compatible. It does, > however, work with modern C/C++ compilers a lot better, is > reenterant, and one can have multiple lex scanners in a program. > Because of that, and the fact we have flex-old, we are not going to > revert the default flex in Debian (and make people revert the changes > they have gone through).
I have not and will not dispute that there are important improvements in the new flex. Reentrancy and cleaner output are certainly wonderful attributes. But why keep the old symlinks when it isn't compatible with the program it is pretending to be? It would be like a fancy new "ls" replacement where the -l option printed a usage message. It would be better to keep the old ls around under the old name, and the new one could be named something else. I have installed flex-old to fix the problem, and it works ok now, but I didn't know to do that until I reported this bug. In fact, I had to read between the lines to figure out it was a flex change and not a bug in the software I was compiling... I had to search the Web for compile errors where one argument to a macro which expects zero. I found no matches for the software I was compiling, but there were matches with people asking questions about compiling software on Debian systems which had recently been upgraded. Nobody seemed to know why it was happening, so I assumed it was a flex problem with the newer version of GCC. I only discovered the documentation about lex/POSIX compatability int hte dpkg output when I submitted the bug report. > Once once has converted ones code over, once can continue to > use the same build system (that calls lex, etc) as one used > to. Indeed, there have been no complaints about the compatibility > links so far, apart from this bug report. I can certainly answer those mailing list posts, though they probably figured it out for themselves or gave up a while back. Some of them might be willing to file bug reports if you would like to have additional ones. > Now, this is admittedly a documentation bug -- flex and POSIX > sections of the docs do need to be rewritten. Ok, sounds good. Thanks, -Ross -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
