On Tue, Feb 17, 2009 at 07:49:33PM +0100, Joerg Sonnenberger wrote:
Point is, release build should never differ whether hg is installed or not.
I absolutely agree, but if by "release build" you mean "a build not in a Mercurial repository" then that is already true of the current version of the patch. If you're not in a Mercurial repository then exactly the same code is compiled whether Mercurial is installed or not.
I know that hg does the check internally, but that also depends on that no other hg binary can be found.
Do you mean that we should worry that there might be an hg in $PATH that is not Mercurial? Is it really necessary to be that cautious? If we must handle host OSes that break our attempts to auto-detect hg then we will have to make the code conditional on a new configure flag (eg. --enable-hg-logging), with the proviso that if you supply the flag then the obligation is on you to ensure that your hg is a properly working Mercurial executable. That's easy to implement. I hope I don't seem obstinate, but I am finding it hard to see the scenario that supposedly breaks the current patch. (Unless the scenario is "hg might not be Mercurial" in which case I have already described what I think is the only sensible fix.) Regards, Jeremy Henty