If anyone wants to muck around, the #Emacs source tree still has a bunch of XPM images, and more or less assumes that -lXpm will work.
It already supports PNG, too. Maybe convert the XPM to PNG?
(bigger project: make emacs use glycin for sandboxed image loading on Linux)
Federico Mena Quintero
in reply to Federico Mena Quintero • • •Things that blew my mind today about Emacs:
* It requires at least librsvg 2.14, which is from 2006, which is when rsvg_handle_get_dimensions() was introduced. That function was deprecated in 2.52, in 2021. If available, Emacs will use all the new APIs. There is #ifdef hell in there.
* src/image.c is 13K lines of code.
* Emacs still uses debbugs.
* The first commit is from 1985, but the first hundreds of commits are weird. It normalizes in 1991.
* And Savannah still uses cgit.
John Regehr
in reply to Federico Mena Quintero • • •it's quite a bit older than 1985 though, right?
my favorite emacs fun fact is that it undumps -- basically breathes life into a core dump in order to avoid doing a large amount of slow initialization work. or at least it used to, I don't follow this...
Federico Mena Quintero
in reply to John Regehr • • •@regehr oh, yeah, it's older. I don't know when the current source tree originated, or even if it went rcs -> cvs -> ??? -> git.
Undumping is so cool! A bunch of lazy initialization bullshit would go away with undumping, for compiled regexes and all that.
Bradley Kuhn
in reply to Federico Mena Quintero • • •re: early commits in Emacs.
I had this side project to normalize the old RCS/CVS history for a lot of the early #GNU programs into Git. I think maybe my retirement hobby will be to pick that up.
Also, There is a set of “lost tapes” — open-reel magnetic tapes that had the 1984-circa 1994 early GNU backups – a goldmine of history. When I worked at FSF, we sent a giant box to a volunteer to bring them onto modern disks, & AFAIK the volunteer ghosted & no one knows where they are. 😕
Federico Mena Quintero
in reply to Bradley Kuhn • • •@bkuhn current tooling to go direct from CVS to git must be incredibly better than what was available years ago, what with the gcc -> git migration.
The two times (I think?) GNOME has changed VCS, from cvs -> svn -> git there was a small, dedicated team of people basically writing custom tooling to do it (or adapting tools to do that), and there are SO MANY SPECIAL CASES. A bunch of repos have funny commits near the beginning, and then they settle down and appear normal.
Bradley Kuhn
in reply to Federico Mena Quintero • • •I haven't looked who the 80s' stuff imported, but I suspect it was particularly bad (as you said) because:
GNU development used SCCS (proprietary), then¹ RCS once it was stable enough, and *then* imported it into CVS from there, which, in turn, was imported to Git. I am pretty sure no one has the backups of the old RCS files (except the lost tapes I mentioned upthread), & to get the truly old data well-imported, we'd need those.
¹There may have been YA RCS-like one inbetween.
Ivan Molodetskikh
in reply to Federico Mena Quintero • • •Federico Mena Quintero
in reply to Ivan Molodetskikh • • •zrzz
in reply to Federico Mena Quintero • • •Federico Mena Quintero
in reply to zrzz • • •Data π🐀
in reply to Federico Mena Quintero • • •kyo.iroiro.party/en/posts/why-…
Why Rewriting Emacs Is Hard
Kyou is kyou is kyou is kyouFederico Mena Quintero
in reply to Data π🐀 • • •