IRC channel logs
2014-10-30.log
back to list of logs
<civodul>alezost: it would be nice if we could mark two generations, press '=', and get a diff of $generation/manifest <civodul>or, even better, the output of (lset-difference ...) <civodul>..[ this was the to-do item of the day, generously offered by yours truly ].. <alezost>civodul: yeah, I had some similar idea, but how to display such difference? Some packages were removed and some were installed *alezost also always wants to pay some time to a civodul's idea of a convenient using of "guix.el" to install emacs packages; but he's always distracted by other things <jmd>How does one do cp -a in guile ? <alezost>civodul: the easiest would be to display the added packages for "=" and removed packages for "C-u =". It would just require a new search-type to be added <bavier>would a diff-style output be appropriate? <bavier>unrelated: I get a "Failed to autoload make-session in (gnutls)" error when building a package with an https origin <civodul>alezost: this format looks good, or it could be more diff like, or a list of added/removed/upgraded packages <alezost>I think diff-style would be good for "guix package" command, but I don't see how it can be used in emacs UI <taylanub>jmd: there's probably no short way to do it. things like that are why I think it would be neat if GNU coreutils, findutils, etc. were made to be libraries. <bavier>alezost: thanks for the pointer. My problem might be slightly different <bavier>the error comes during a 'guix build' rather than 'guix download' <bavier>alezost: I see. Yes, reverting that seems to resolve my problem. <bavier>but I see that there's much discussion around the versioning question. <alezost>mark_weaver: civodul: I think the current situation with gnutls is the worst. Either my commit should be reverted, or a path in "guix/download.scm" should be adjusted to "/share/guile/site/2.0". What do we do? <civodul>alezost: sorry, i was waiting for an answer from mark_weaver, i think; i've just pushed the simple fix for that <civodul>we should keep discussing the more general issue, of course <bavier>hmm... patch-shebangs doesn't recognize "# !/bin/sh"? <taylanub>bavier: not even sure if most unixen will allow that? <taylanub>if GNU+Linux does then I guess Guix should too, though I'd fix any file doing that. no whitespace is conventional, and the only alternative I've ever heard of is a space after the #! <bavier>taylanub: I can execute the script outside the build environment, presumably because the /bin/sh is available there, but not within the build chroot, because patch-shebang doesn't patch /bin/sh <civodul>bavier: valid shebangs match "#! ?/.*" <bavier>civodul: do we just patch strange shebangs at the package level then? <bavier>and why would the "# !" shebang'd script execute outside guix? <civodul>bavier: dunno, is "# !" actually valid? <civodul>and the default, in the absence of shebang, is to use /bin/sh <civodul>yes, if you look at fs/binfmt_script.c in Linux, it's pretty clear: it has to start with exactly "#!" <viric>civodul: the default if called from sh <viric>I mean, the shell tries to run it if exec fails, no? <civodul>libc doesn't do anything fancy like this <civodul>it just passes on ENOEXEC, apparently <bavier>blech, turns out better to not even use their autogen.sh script <bavier>I sent some patches to guix-devel with get send-email this time, but it doesn't look like they've made it yet... <alezost>bavier: do you use gmane? There are some problems with gmane today; at least I receive "No more unread articles" all day <alezost>bavier: ah, ok, I thought you used gnus <alezost>bavier: hm, strange, I think there were more messages today than in that list