➜ tlmgr update --self --all
tlmgr: package repository ftp.fau.de/ctan/systems/texlive/tlnet (verified)
tlmgr: saving backups to /opt/paulo/applications/texlive/2021/tlpkg/backups
tlmgr: no self-updates for tlmgr available
/opt/texbin/tlmgr: create(tlpkg/tlpobj/tex4ht.tlpobj) failed: Permission denied at /opt/paulo/applications/texlive/2021/tlpkg/TeXLive/TLPOBJ.pm line 625.
@Rmano Thanks. :) I could not reproduce in my other machines (three Linux installs), dunno what happened. But it's so strange because for whatever reason root took ownership of lots of files!
@PauloCereda I'm a bit disappointed by the favicon of the matrix website -- so much potential for matrix jokes and what do we get? A boring square with some circle thingy
@UlrikeFischer `£ would work just the unicode number for £ I suppose. although I'm not totally convinced fontspec is right to be messing with the classical tfm math setup at all
@UlrikeFischer I have reported this previously at texlive (not that they changed it) but is this worth reporting at miktex or is there a user error here in the setup?
I have found it, I think. It is located in my personal config folder (C:\Users\tethe\AppData\Roaming\MiKTeX\fontconfig\config). There is a file called localfonts.conf. If I delete the line <dir>C:\Programs\MiKTeX\fonts\type1</dir> from it, the offending file seems to compile. I just need to take care NOT to refresh font map files in MiKTeX console, as it seems to write back the deleted type 1 line. — tethered.sun23 mins ago
@DavidCarlisle I don't think that it is an user error, I have type1 lines there too. I will report it. Imho they should remove them, it is only a source of problems.
@PhelypeOleinik since I don't want to nag Chris more than necessary, I'll write here: l3benchmark uses an arbitrary operation as measurement for one op (IIRC it uses one \def to empty, but am not sure about this, been a while since I looked at its code). And I never was pro-unprotected. I just stated the objective facts without any indication of what I'd prefer. As a matter of fact, I'd agree to your view, impact is negligible, but the correct "standard" would be to use \protected.
@PhelypeOleinik ok, took a look, \tl_gset:Nn \g__benchmark_code_tl { \int_gadd:Nn \g__benchmark_duration_int { 0 } } is the code that is run as one op (that or 1e-16 as a minimum value).
@Skillmon Oh, I see, it uses a reference time to count ops. Good to know, thanks!
@Skillmon Headline statistics like “\protected macros are 6% slower” tend to stick, even if they are not quite realistic (no actual document expands \prg_do_nothing: thousands of times in a row :), which makes arguing in favour harder, so I tend to avoid those
@JosephWright is this just a rename do you know? xcolor2.pdf (which is distributed as pdf on ctan) fails with ! LaTeX Error: File `fvrb-ex.sty' not found, texlive has a fancyvrb-ex.sty I'll see if that works...
seems to work, I wonder if we really need to build xcolor2.pdf for ctan or drop it @JosephWright @UlrikeFischer ?
@JosephWright is there an easy way to make xcolor.pdf with pdflatex and xcolor2.pdf with latex/dvips/ps2pdf -ALLOWTRANSPARENCY or does that need a custom typeset lua function?
@UlrikeFischer ^ but now (or probably tomorrow) need to get l3build to do that..
@UlrikeFischer did you get chance to look at the actual change of using \@backslashchar a bit scary doing an xcolor release it'll be embarassing if we break it:-)
@UlrikeFischer yes I looked at the test files that are extracted but they seem to be big tables of colours I suppose we could convert them to pdf based tests