M, > changes in the system include files which affected the result?
I thought that too, so I put features.h.orig back ontop of features.h. And it still kept compiling. So I thought, I must have modified the original back too. No problem, I'm compiling glibc, I have the source lying around. diff'd features.h against 2.2.4-1 and 2.2.3-9 and came back with nothing... the file is original. Which really wigged me out :) > The most confusing part of this is that everything works fine on i386, and > features.h is exactly the same between 2.2.4-2/i386 (my system) and > 2.2.3-9/hppa (paer). Same goes for netinet/udp.h. sys/cdefs.h only differs > inside a comment block. So I think I'm inclined to blame g++, and I'm going > to reassign the bug there until I figure out what else to do. Now that you mention it. It probably looks more and more like a g++ bug. I can't compile that test program with gcc or g++. However, since my results _seem_ to be tainted, I'll run through them again this morning. I'll get back to you in the afternoon. > Which version of libc6-dev were you running on your test system? I had been > using paer, which has 2.2.3-9 installed. I was using the latest apt-get dist-upgrade'd version, which is 2.2.3-9. Incidentally the last version that compiles on hppa. > If you don't mind publicizing it, could you send a copy of your analysis to > [EMAIL PROTECTED] I'm not copying this message there because I'm > quoting your text; if you don't mind, I'd like to send this message there as > well, to help others in tracking down the problem Send away! I'll do a proper bug report when I get into the office. --- Bug Test: - libc version - gcc version - ld version Test basic program with _BSD_SOURCE (gcc and g++) Test with #undef _BSD_SOURCE (gcc and g++) Outline previous description. Test it on 712/60 and 715/50. Record results. Email to [EMAIL PROTECTED] --- Cheers, Carlos.
