On Feb 12 2014, at 09:31 , Henry Jen <henry....@oracle.com> wrote: > On 02/12/2014 07:11 AM, Magnus Ihse Bursie wrote: >> 12 feb 2014 kl. 14:55 skrev Erik Joelsson <erik.joels...@oracle.com>: >> >>> First I was disappointed to lose the configure-arguments file, which I >>> sometimes look in to see how configure was run, but I can just as well look >>> in spec.gmk. >> >> Yes, that was my own reaction too. :-) "Hey, I can't remove that". :) I'm >> not even sure why I initially didn't think of putting it in the spec.gmk >> instead if a separate file... >> > > Usually I checked at config.log, I didn't even know this file exist.
ditto. I look at config.log or config.status. > >>> >>> Have you tried running with complex arguments, like >>> --with-extra-cflags="-flag1 -flag2"? >> >> No. It will probably fail. :( Maybe we can detect this and warn/refuse to >> run, but I don't even want to start thinking how we could support that in >> the reconfigure target. It's probably not worth it. This worries me. If someone does use complicated options that would fail on re-run perhaps it's best to fail immediately rather than leave a mysterious (possibly silent) failure for some later point. Either fully support it or don't support it at all. Imagine --with-extra-cflags="-flag1 -q" If the -q gets turned into a configure option on reconfigure what havoc might this wreak? > Dealing the quote is not fun, I agree it's probably not worth it. > > Cheers, > Henry
