Greg Schafer wrote: > I said it looks completely bogus, coz to me, it does look completely > bogus! :-/ It might work for the task at hand, I dunno. But I haven't > tested it, and I'm not likely to test it either, because I do not believe > in using wrappers for toolchain components. IMHO wrapping toolchain > components is fundamentally evil. Time and again, it has proved to be a > risky thing to do. The most notable example I can think of came up on the > Glibc lists a while back whereby some folks were creating a wrapper around > GCC to force it to build as "gcc -march=i686". (Of course these days > there are proper supported ways to achieve the same effect). This practice > will completely break a build of Glibc for non-obvious reasons. It > resulted in bogus "my build failed" type posts to the Glibc lists and > bogus bug reports thus wasting valuable Glibc developer time.
Thanks, this is good to know. > I suppose the real point I was trying to make is that you were very shoddy > in making these changes. You obviously hadn't tested it, there was no > consensus on the list, and worst of all, you basically went ahead and > invalidated many man hours of Dan's time working on ICA stuff, which he > was doing essentially for your benefit! (you being the driver of the new > build order). The whole thing is just very frustrating for followers of > LFS development. Quite frankly, I can see why Randy gets all annoyed > occasionally :-) Yes. I thought of these things. Unfortunately, only after the changes had been made. I had followed a general course plan that Matt had laid out, assuming all the wrinkles had been smoothed out ahead of me. I also assumed that the community had discussed it enough to form a decision. And well, assumptions... On the other hand, as you said, the toolchain was broken, in that chapter 6 binutils and gcc linked against the wrong glibc. I would think that is a necessary fix, (regardless of the recent ICA results based on it), and also that change of itself would require another round of ICA testing. The current setup should work, (at least temporarily), as it closely resembles what you do. The only items being built by the gcc wrapper should be binutils and GCC. I am working hard at redeeming myself here by testing (yes testing!) a setup that whould allow us to both drop the wrapper, any -B flags, and skip the retaining of the binutils dirs. > Anyhoo, not to worry. Hopefully lessons have been learnt and the ship is > back on course. Indeed. And hopefully no real damage done. -- JH -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
