On Friday 24 August 2001 10:50, Tom Badran wrote: > > Basically, our 2.96 is the occupation of a number that was abandoned by > > the gcc team. That abandonment came about because another distro > > occupied the number. The two came from the CVS development tree at very > > different times and do not really resemble each other very much, yet the > > binaries are mostly compatible. > > > > It may be built from many patches out of the cvs tree, but it > > consistently produces reliable code. Most of what people see as a faw in > > it is that a lot of sloppy code written for 2.95 and earlier expected the > > compiler to load standard headers by default. 2.96 needs the explicit > > #include statements. > > Ok, the biggest problem i have with it is that trying to build a cross > compiler using the gnu gcc 2.95.3 sources seems to have problems, would i > be better to compile gcc/binutils using kgcc? (which i believe is egcs). > Maybe most of the problems i have are due to sloppy code as you say, and > therefore i would ask which libraries specificallly are no longer included > as default. > > Oh, and are you saying that your numbering gcc as 2.96 has nothing to do > with th gnu version numbering system? As i thought their 2.96 series were > just 3.0 betas (in which case i am a little mixed up as i used to use a > standard gnu version of 2.96). > The gcc team officially abandoned the number after RH issued it in their 7.0 release. We started using a pruning from the CVS tree sometime later (after others had fixed many of the bugs :-D ) and continued to add patches. I don't know what kgcc is, but I know most people have had trouble building our kernels with it. In fact, someone flamed us on this list because he had to do a make mrproper and because kgcc (which we never issued) built a kernel with an endbase address too big. He asked what EDBA was and I gave him a reference to the authority on it (a LILO manual) and he accused me of sending him on a treasure hunt. Basically no headers are included by default, and there are some other strictnesses, but the number of "internal compiler error" messages we have had are tiny. I believe three, all from packages we never saw. How we are expected to diagnose without seeing the input to the compiler is beyond me, but folks report the bug that way. Anyway, I hope this answers your question. This is basically a unique compiler and we will return to the mainstream compilers as soon as we have one that does as well. Civileme > Thanks > > Tom ---------------------------------------- Content-Type: text/plain; charset="iso-8859-1"; name="message.footer" Content-Transfer-Encoding: 8bit Content-Description: ----------------------------------------
Want to buy your Pack or Services from MandrakeSoft? Go to http://wwww.mandrakestore.com
