On Fri, Oct 16, 2026 at 11:37:12PM -0700, Thomas Frohwein wrote: > On Mon, 5 Oct 2026 16:42:56 +0100 > Stuart Henderson <[email protected]> wrote: > > > On 2026/10/05 05:24, Thomas Frohwein wrote: > > > - devel/jj-gh: non-standard GH tagnames (jj-gh-v0.3.1); therefore > > > justifiable to require this to set WRKDIST manually. (There may > > > be a smarter regex for *_TAG_MOD.github that can avoid this.) > > > > > ok? > > > > this is not an unusual naming format - while it's uncommon for > > DIST_TUPLE you'll find 30-40 or so if you look at ports with > > GH_TAGNAME so would be good to handle it sooner rather than later > > (setting WRKDIST=${WRKDIR}/${DISTNAME} by hand isn't something that > > easily comes to mind..) > > It took a bigger rewrite and some testing to arrive at what now avoids > the previous simplification that led to the regression with jj-gh and > others with a TAGNAME like abc-v0.1.2. Diff below, and ports that need > patches to deal with the changes (for good reasons) attached. > > Reminder about the rationale for update: lack of DIST_TUPLE flexibility > is leading to an increasing number of workarounds in dist-tuple.port.mk > and individual ports. Examples include custom > MODDIST-TUPLE_post-extract for codeberg so far (and videolan has been > proposed), and a significant number of DIST_TUPLE users having to set > WRKDIST to make up for the limitations. > Ideally, DIST_TUPLE will arrive at: 1) all configuration for upstream > sites is done in the variables in db/dist-tuple.pattern, and 2) no > workarounds are needed in individual ports for the pertinent variables > like DISTNAME or WRKDIST. > > The main thing to know for those who want to add new sites/templates > into dist-tuple.pattern is that TEMPLATE_EXTRACT_DIR.${template}.{hash > or tag} tell dist-tuple.port.mk what directory the upstream distfile > extracts to (relative path). Here, '<mod_id>' can be used in > combination with setting ID_MOD.${template}.{hash or tag} which is the > 'id' part of DIST_TUPLE (4th element of the DIST_TUPLE string) modified > by the pertinent ID_MOD.x.y. This way, the flaw of the previous version > that such modification was applied to the whole string for WRKDIST is > now avoided. > > As a result, workarounds for codeberg (custom > MODDIST-TUPLE_post-extract and setting WRKDIST) are not needed anymore. > > This diff also fixes the issue with having to set WKRDIST in case of > gitlab with commit hashes (raised by ian in [1]). If/when the update in > this or a future version goes in, we should weed out the then > unnecessary WRKDIST lines. > > I can comment on the internals of the new dist-tuple.port.mk structure > upon request. There may be opportunities to refine a few things, like > that there is quite a bit of duplication between the # tag mode and # > hash mode sections that I couldn't find a way to improve sanely. > > I ran this diff with the attached patches through a complete bulk build > and no errors attributable to this. Rationale for the diffs with some > explanation why I think they are a necessary step now below in response > to sthen's useful WRKDIST before-after comparison. > > ok?
Please also write the corresponding man page snippet documenting the new TEMPLATE_EXTRACT_DIR variable. And what needs to be user-visible in there. Either your feature is good and it will be obvious, or you'll find that while trying to explain it in a manpage, it can use some love.
