On 12/27/18 6:39 PM, enh wrote: > On Thu, Dec 27, 2018 at 12:18 PM Rob Landley <[email protected]> wrote: >> >> On 12/27/18 12:46 PM, enh wrote: >>> meh, BSD-based regex implementations seem to think that the empty >>> regex is an error anyway. (i only skimmed it, but i think POSIX >>> agrees: >>> http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap09.html) >> >> Earlier versions I was catching the empty regex and manually matching the >> whole >> thing, I can do that again. :) >> >> (Some build script somewhere cared, I forget which since I haven't gotten the >> LFS/BLFS build environment recreated yet...) > > really? isn't this just a pathologically stupid way to say `cat`?
Back circa http://lists.busybox.net/pipermail/busybox/2004-May/011538.html one of the very first problems I hit was autoconf using awk with an empty string as a cat synonym, and at the time busybox awk hung. (What they did was have a $VARIABLE accumulating the transforms they wanted applied to some header file, and the common case for Linux was no changes, and at the end they'd call cat with the $VARIABLE and it would just copy the file unmodified for Linux. It makes a certain amount of sense in context, but means all sorts of weird behavior winds up being depended on. :P) Rob P.S. You want pathologically stupid, commit 32b3587af261 was fun to debug. _______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
