At Monday 26 July 2010, Ralf Wildenhues wrote: > Hello Stefano, Hi Ralf. How did the GHM go?
> * Stefano Lattarini wrote on Mon, Jul 26, 2010 at 01:58:08PM CEST: > > This patch substitutes (almost) all the sed/grep maintainer > > syntax checks in `Makefile.am' with a new perl script > > `maintcheck.pl'. > > Why? Honest question. perl regex isn't necessarily more readable > than sed, At least in this case, they are not. But the main reason for this new perl implementation was to allow easy whitelisting and extensibility of checks. With checks as heuristic as ours, I think that the possibility to easily whitelist lines and files is essential. > in fact I find some of the constructs you introduced > decidedly less readable. True. But most of them are also more precise. > Also, testing in the presence of VPATH builds is a requirement. I heartily agree. That's why I added a "RFC": the patch is by no means definitive or complete. > I think a reorganization of maint checks, *if at all* done, should > be done unified to/across gnulib and coreutils' infrastructure, > rather than yet another reimplementation. The new maintcheck.pl script, with some small refactoring to introduce external config files and support for VPATH builds, might IMHO work as a general line-oriented syntax checker. > I don't really see the need though, the existing tests are not > broken, are they? IMHO they are, since they lack an easy unobtrusive way to whitelist false positives (also, some of them might not work with non-GNU grep, but that's just a minor nit). Regards, Stefano
