On Thu, Oct 15, 2026 at 08:38:47 -0700, Alistair Crooks wrote: > On Wed, 14 Oct 2026 at 16:36, Valery Ushakov <[email protected]> wrote: > > > Shall we perhaps make it (the new code) our usr.bin/grep? > > I've always resisted this in the past, as the Perl escapes are just the > ones that I use myself, they're not the complete set, and might make this > whole thing "non standard", FSVO standard. It's also why I gave it an > egotistical fluffery name, rather than calling it some form of grep :) > > But, at the same time, I see egrep hasn't changed in 9 years since I wrote > the README, and so, right now, I definitely wouldn't expose any regular > expressions to the internet without making sure they couldn't be abused. We > also don't have any wchar_t expressions in base, and I can imagine people > getting confused by some things in the library, like the virtual > instructions it can dump, or the execution tracing. > > I might try a short experiment to use the new code in a build.sh run, and > see what happens - that's where previous attempts (with tre) have failed, > mainly due to the distinction between basic and extended regexps, IIRC.
Much appreciated. The grep situation is currently very sad. Out gpl2 grep is _very_ old and it's utf-8 performance sucks massively. A modern amd64 machine might hide that, but even on a 1GHz g4 grep -r /usr/include is excruciatingly slow. Unfortunately, the fixes for that are in the gpl3 version only. Thanks. -uwe
