> On Oct 29, 2014, at 10:44, Mark Johnston <markj_at_freebsd.org> wrote: > >> On Wed, Oct 29, 2014 at 09:18:58AM -0700, Garrett Cooper wrote: >> >>> On Oct 29, 2014, at 07:42, Ian Lepore <ian_at_FreeBSD.org> wrote: >> >> ... >> >>> Why are you the only person in the world who has had to do this? >> >> They're not the only one. $work is running into an issue where there's a mismatch between the host and build version which is generating corrupt ctf sections. >> >> I thought it had been fixed on CURRENT recently, but the converse might be true--it might have been broken because it's no longer being built as part of the bootstrap tools process. > > Are you referring to r266567? I'm looking at whether we could > unconditionally add ctf* to the bootstrap tools. That would address the > OP's problem I think, as well as the corruption issue. > >> >> My gut says that this section is prematurely optimizing it out of the build, because it's assumed that the host tool will always be binary compatible with the build tool: >> https://svnweb.freebsd.org/base/head/Makefile.inc1?annotate=273755#l1270 Yes. ctfconvert always needs to be built with the build tree version instead of relying on the host version because using the host version will break if/when the ctf format between the build and host machine changes. Some care might need to be taken with the library dependencies for ctfconvert, but libelf is a relatively simple library from what I remember... Thank you!Received on Wed Oct 29 2014 - 17:16:28 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:53 UTC