On Mon, 13 Jan 2014 at 14:24:58, Martti Kühne wrote: > It may sound like two dumb or unrelated ideas, but I figure they would > become relevant to the discussion at one point anyway. So, let's look > at the question we're avoiding all the time: What is the problem with > PKGBUILDs? That it's not limited to a subset of sh?
It cannot be parsed without full execution. Theoretically, a PKGBUILD can contain something like pkgname="$(echo foo | sed 's/o/a/' | awk '{ print "b", $0, "r" }' | tr -d ' fo')" and something even fancier. > > There seem to be regularly efforts appearing and disappearing to clean > that mess up on the initial side of the problem, which could render > this .AURINFO / .SRCINFO ballet unnecessary. > > How about creating a parseable and functionally equivalent > "PKGBUILD.gen" from PKGBUILD? So, you're basically evaluating all PKGBUILD variables on the client side, then print them into a file called "PKGBUILD.gen"? That's exactly what .AURINFO is, except it has another file name. > > cheers! > mar77i