On 3/1/06, Jason Stevens <[EMAIL PROTECTED]> wrote: > I've been following the list for about three months now, and during that > time I've built three versions of HLFS: 20060101, 20060108, and > 20060220. I've dutifully rebuilt the toolchain for each version (though > not for multiple builds within a version), and it's led me to an > (obvious) realization: I should be able to reuse an "older" toolchain > to build a newer version of the source.
Interesting thought, and though I know little at all about toolchains outside what i've learned reading these lists, it sounds reasonable for a short timeframe, though the changes to the versions within the CH5 toolchain could (should this be would?) affect the build of gcc and glibc, meaning the entirety of the build could have unexpected problems. > As I read the HLFS book, during Chapter 6 (building the host system), > the path is set to use the tools just built over those in the toolchain > as soon as they are available (the +h). Since almost the first thing we > do in Ch 6 is rebuild glibc and gcc, those should then be used for > building any later code. The only exception is binutils, but since we > don't patch that at all in the toolchain, I gather it should be fine. true, though these are the toolchain for the final system, the CH 5 toolchain is built separate from everything else to seclude it from the host, and any problems the host could introduce, as well as in HLFS, to provide the increased security related factors, if problems are found in older toolchain packages, or the HLFS specific factors change, the old toolchain becomes completely obsolete. > So... Is my conclusion that I can reuse an older toolchain (including in > that the binutils-build dir that we need to keep for adjusting the > toolchain later) correct? And if so, how would I know when an older > toolchain is "expired" for a given version of the book? Keep an eye on changes to the toolchain packages and the reasons, as well as patches to them. > The only way I can think of to actually test the hypothesis is to build > a system with an older toolchain and the current one, and compare them. > Is there some better way to do this than to recursively test for same > structure and file checksums? Hm, even that would be more tedious than it initially sounds simply because of the problems introduced by a timestamp being embedded somewhere obscure or some other little thing. > Thanks, > > -jps Hopefully my minimal understanding of the matter helps at least a little, and isn't entirely incorrect any any of it's parts, good luck ... one thing i've really learned from *LFS, only way to know for sure if something works is to test it. -- Poison [BLX] Joshua M. Murphy -- http://linuxfromscratch.org/mailman/listinfo/hlfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
