@sandro@hfaust@sun With most of ripgrep performance being from the research done by the author of the_silver_searcher aka ag, and either implementation would likely make your whole system a *lot* faster if there would be a grep(1) compatible interface. And by that I mean mostly posix standard, with maybe few bits that are common between GNU grep and BSDs grep.
At least I don't think there's any need for bug-for-bug compatibility, that's what Enterprise distros try to do, upstreams meanwhile tend to fix their bugs and occasionally make breaking changes to APIs.
Also rewritting things in bug-to-bug compatibility is no fun. Writing a tool that replaces the existing one with useful new features and at the same time being mostly memory safe and sometimes even was faster is a way more motivating combination. See ripgrep.
@sandro@hfaust@sun Well that part of autoconf is designed broken, like pretty much all of autoconf.
I don't think I've had a year where it didn't have a serious bug, even just compilers getting 2 digits in their major version broke it. And so each time more and more autoreconf gets added to packages, typically coming with patches so autoreconf would actually work, or even patches to ./configure directly because it just wouldn't work.
@lanodan@hfaust@sun Using ripgrep as grep in userland mostly works with the exception of -E. Using it systemwide will probably unveil many edge cases in especially in shell script monsters like autoconf who check eg the copyright header of mkdir if it supports -p.