IRC channel logs


back to list of logs

<jmd>mutt -y
<civodul>Hello Guix!
<civodul>jmd: does the fltk patch depend on the libxrender patch?
<jmd>If the libxft does not propagate libxrender then all packages which use libxft will have to also list libxrender as an input.
<jmd>So far as I can see, libxft cannot be used without libxrender
<civodul>because it linking otherwise gives an "unresolved symbol" error, right?
<civodul>ok, i see
<jmd>Its not visible to the user. But thats what pkg-config gets, and then pkg-config thinks that libxft is not installed.
<civodul>jmd: but xrender.pc does not "Require:" Xft
<civodul>so i don't get it
<civodul>maybe it's just that fltk needs libxft as an input, no?
<jmd>Other way around.
<jmd>xft.pc requires xrender
<civodul>ah, right
<jmd>or it should do. I didn't check.
<civodul>it does
<civodul>lemme see
<jmd>fltk already has libxft as an input. Adding libxrender as an input would also work. But why do it that way?
<civodul>ok, so i'm applying the xft patch
<civodul>yes, you're right
<jmd>Right now I'm struggling with a wierd problem.
<jmd>I have a makefile which has this rule:
<jmd>libaegis/aer/ libaegis/aer/report.gen.h: \\
<jmd> libaegis/aer/report.y
<jmd> @echo Expect 2 reduce/reduce conflicts.
<jmd> $(YACC) -d libaegis/aer/report.y
<jmd> sed -e 's/[yY][yY]/aer_report_/g' -e '/<stdio.h>/d' -e \\
<jmd> '/<stdlib.h>/d' -e '/<stddef.h>/d' > \\
<jmd> libaegis/aer/
<jmd> sed -e 's/[yY][yY]/aer_report_/g' -e \\
<jmd> 's/Y_TAB_H/aer_report_TAB_H/g' > \\
<jmd> libaegis/aer/report.gen.h
<jmd> rm -f
<jmd>In the chroot YACC returns 0 but does not create the (it does create the
<jmd>Consequently sed fails.
<civodul>what's the value of YACC in config.log?
<jmd>bison -y
<civodul>we have the latest Bison
<civodul>so it could be that for some reason this package expects an older one
<jmd>Oddly it works fine when I run make from /tmp/nix-xxxx
<civodul>after sourcing 'environment-variables'?
<civodul>or could it be a parallel make issue?
<civodul>you can try #:parallel-build? #f
<jmd>Oh perhaps you are right.
<jmd>if the is getting deleted by another rule
<jmd> is a global!!
<jmd>Where does the #:parallel-build? #f go ?
<jmd>In (arguments) I presume
<jmd>That would also explain why it worked for me once.
<civodul>well, grep parallel-build *.scm :-)
<jmd>I think it has fixed the problem. Thanks for catching that one!
<jmd>I'm surprised Peter Miller made that mistake.
<civodul>the Peter Miller of "Recursive Make..."?
<jmd>Yes. Same guy.
<civodul>indeed, impressive :-)
<jmd>guix package doesn't accept the --target= option. How can I install packages targeted at other systems?
<civodul>jmd: that's a weird thing to do ;-), but you can do: guix package -i $(guix build --target=foo ...)
<jmd>Not all that wierd. I have my fileserver mounted from machines with different architectures.
<jmd>It would be good to allow: guix package -i --target=foo
<jmd>But it should then default to the profile_dir .guix_profile_foo
<civodul>actually, i was thinking that when we support several kernels, then it would be great if users could end up have ~/.guix-profile point to a version of the profile for the current kernel
<civodul>so you could boot into the Hurd, or Linux-Libre, and things would just work
<jmd>sounds reasonable
<civodul>BTW, i should probably put a message on saying that Guix will be at FOSDEM
<civodul>so we have more chances to meet interested people
<jmd>Sound like a good idea.
<civodul>done :-)
<jmd>I've got some tests which are writing to $HOME is it safe to set it in the test phase?
<civodul>yes, you'll have to set $HOME to a writable directory
<jmd>Can I use a glob pattern in substitute* ?
<civodul>jmd: nope
<civodul>the first argument can be either a file name, or a list of file names
<civodul>so you can use the 'find-files' procedure to list files matching a given regexp (recursively)
<jmd>Great, thanks.
<davexunit>hey #guix, I want to write a package for guile-sdl. there's a module for guile and module for sdl. where should it go?
<davexunit>I'm leaning towards sdl
<civodul>davexunit: i'd say guile-sdl, but either way is fine
<civodul>sorry, guile.scm
<davexunit>civodul: roger that.