On 24 July 2014 02:43, Assaf Gordon <assafgor...@gmail.com> wrote: > Attached is my attempt at such patch, based on your suggestion.
Hi Assaf, That's great. The weekend following when I wrote the reply I polished a bit of the 'patch' I originally sent, that was completely broken, that end up to be remarkably similar with the patch in your latest mail. But not exactly the same, and I must admit yours is more elegant. The greatest difference is that you skip manual page generation when it's pointless. > Few differences from your suggestion: > 1. If building from tarball (as opposed to from .git), I've changed the > build to never generate, and never clean the "hello.1" manpage. The downside > is that if a developer works from a tarball and changes the help screen, the > manpage will be stale. > This is the "BUILD_FROM_GIT" in Makefile.am . > I think that's reasonable, but I could be wrong. I think skipping manual generation when doing tarball build is good strategy. The manual is a frozen, like the code, and there really should not be reason spend time regenerating it (or them, depending on project). > 2. If building from GIT and cross-compiling, don't add a rule to the > makefile to generate the man-page (this is the "!GIT_CROSS_COMPILING" in > Makefile.am). Ack. > 3. I've read somewhere (can't find the link now) that AM_CONDITIONAL should > not be nested - so it is outside the "AS_IF". Yes, that was one of the bits not quite perfect in initial mock version of the change. > 4. To detect the ".git" directory, I've used "$srcdir/.git" (not just > ".git") - as this could be problematic in out of tree builds? All I can say the 'make distcheck' works, and $srcdir/ sounds like right thing to do, so no objections. > I've tested it on several systems (including cross-compiling), but more > testing is always good. Excellent. > Comments are welcomed, None other than thank you for contribution. The change is merged. -- Sami Kerola http://www.iki.fi/kerolasa/