Jamie Lokier wrote: > Detlef Vollmann wrote: > > IMHO, a toolchain that requires a specific location in the host > > filesystem is broken anyway. > > All the prebuilt ones I've seen do. It would be great if they didn't, > but they do. As I already wrote, all toolchains built by crosstool location independent. > > > > A pre-built binary GCC 2.7.2 distribution is likely to always function > > > correctly on future hosts where it will run, and always produce the > > > same code, however. > > That pre-built toolchain is probably dynamically linked against > > a C runtime library that you don't have around any more... > > C runtime compatibility is quite good. In practice I've not seen this > kind of problem, even with the the most up to date C experimental > runtime library on GNU/Linux and a toolchain that is dated from 2001 > and built on a different GNU/Linux distribution. I use the same > prebuilt toolchains on several totally different flavours of > GNU/Linux, some very old and some very up to date. Yes, I believe you. But you wrote you want to produce the "same" code. And as libraries as well as compilers are generally not bug-compatible across different versions. So newer versions of compilers or old compilers run with newer libraries sometimes produce better code.
> If it didn't work, I'd recommend virtual machines to run the old > toolchains... :-) But it seems to be fine. We always rebuild our toolchain when we're required to rebuild old versions, and we never hit a problem due to that practise. And several times we had to rebuild the toolchain anyway, as it turned out that some bug was caused by a problem of the toolchain. So the bottom line is, that if you want a byte to byte identical version of your old release, you simply use your binary archive. And if you want to change something, you can as well rebuild your old toolchain. Detlef -- Detlef Vollmann vollmann engineering gmbh Linux and C++ for Embedded Systems http://www.vollmann.ch/ _______________________________________________ uClinux-dev mailing list [email protected] http://mailman.uclinux.org/mailman/listinfo/uclinux-dev This message was resent by [email protected] To unsubscribe see: http://mailman.uclinux.org/mailman/options/uclinux-dev
