On Sunday 13 July 2008 23:59, Poly-p man wrote: > uclibc's regexp code is 53k by itself. > > Since things like expr, awk, etc. need it, I was wondering if it was > even feasible to consider writing our own miniature, > just-what-we-need-and-no-more regexp code, possibly implementing that as > an option or whatever. > > I could do some work on this, but I'd like some input from the mailing > list first.
Yes, sure, if you want to write smaller regex matcher, by all means, do it. Bernhard's comments make sense - instead of writing regex for busybox, you can just reduce/rewrite uclibc's regex. This way your code will be used by more people. uclibc has portions of code which were taken gron glibc (and elsewhere) without thorough code size shrink, there are some interesting possibilities. If you see reasons why writing busybox specific regex matcher is beneficial, please let us know. For example, when I looked at crypto code, I noticed that good chunk of code is used by generic des encoding, which was not actually used by busybox. (busybox only needs more limited interface, just enough to support /etc/passwd style crypted passwd). Using this fact, busybox can get away with 15% reduction of crypto code size. But this may not play out this way for regex. You need to take a look to check it. -- vda _______________________________________________ busybox mailing list [email protected] http://busybox.net/cgi-bin/mailman/listinfo/busybox
