Hi, Here's a new version of the patch. The only change is: I modified the tests (tests/graph.scm and tests/profiles.scm) so that they also use the package-full-name procedure's new #:delimiter keyword argument when necessary. I intended to do that from the start, but I forgot. Alex Kost writes: > Chris Marusich (2018-04-09 23:38 -0700) wrote: > > [...] >> Because I changed the package-full-name procedure, this also fixes the >> package names in other places, such as graphs produced by "guix graph". > > I recall there was some discussion about 'package-full-name' in the past > but I remember nothing about it :-) > > Well, maybe it is not relevant at all, but I've found there was a > proposition to use "@" in 'package-full-name' and it is still not used > there. Look at: > > http://lists.gnu.org/archive/html/guix-devel/2016-04/msg01141.html Yes, we need to be careful. The most obvious problem I encountered was that the Guix daemon prohibits "@" in store item names (which I learned only after attempting to change the delimiter everywhere and then running some tests). However, my patch does not put the "@" symbol into store item names. In fact, to be extra clear, I've added a comment to the package->bag procedure to explain why we can't use "@" there. I think the risk is fairly low that something in Guix itself will malfunction if we make this change, since I've looked at the call sites for package-full-name (in the guix Git repo only), I've run "make check" successfully, and I've verified manually that a handful of guix commands still function correctly. But you never really know until you try. We should probably still check in some other places before changing the default delimiter from "-" to "@". For example, we should probably check at least the following places: * Hydra * Cuirass * Emacs-Guix Where else should we check? If it's too risky, we can always make the default delimiter "-" and only change it to "@" in those places where we really need to (e.g., the package->recutils procedure in (guix ui), to fix this specific bug report). What do you think? -- Chris