Created attachment 369314 [details] build.log Before revision, bash-4.1* built fine. After revision, bash-4.2* failed to build. Attached build.log
This is probably because bash uses USE=readline in more than one way: > econf \ > --with-installed-readline=. \ > --with-curses \ > $(use_with afs) \ > $(use_enable net net-redirections) \ > --disable-profiling \ > $(use_enable mem-scramble) \ > $(use_with mem-scramble bash-malloc) \ > $(use_enable readline) \ > $(use_enable readline history) \ > $(use_enable readline bang-history) \ > "${myconf[@]}" So combining EXTRA_ECONF=--disable-readline and USE=readline (which is forced in profiles/base/package.use.force) cannot possibly work right. Adding --disable-history and --disable-bang-history MIGHT work, but note that USE=readline is forced for a reason (portage needs the functionality to work right).
Yes, but it also only needs this for stuff that doesn't hurt during the bootstrap. So if those two --disable flags work, then I'm fine using that, but if messing around with the profile fits in the current hacks then that would work for me too.
(In reply to Fabian Groffen from comment #2) > bootstrap. So if those two --disable flags work, then I'm fine using that, It allows me to get past this issue.
in fef4efe7f3df
I see, prefix portage has patched away the need for bash[readline]. Maybe the nicer solution is to unforce bash[readline] in the prefix profile then?
I'm not aware of any hacks we made, did we? (Or did Zac do, and forgot to lift the mask again?)
bin/save-ebuild-env.sh, line 89. I don't know who put it there.
That's flagged as a prefix hack, so must be me. Guess clearing the mask is the way to go then.
This should be fixed with tomorrows snapshot