webrev -p is pretty much not possible to support reliably over Hg (for the same reason it's not entirely foolproof with TW).
However, at present, it outright doesn't work (errors from Hg, because the parent (webrev) isn't a workspace). Could we teach webrev to fall back to the unknown SCM logic at that point? (I haven't looked, but it seems plausible, for some reason that didn't occur to me when we were looking at this previously). Beyond that, however, I can't even currently find a way to convince it that it really wants to create an incremental using the "unknown" type, and CODEMGR_PARENT/CODEMGR_WS set to the raw_files/new of the two webrevs in question. That seems like a bug, unless I'm missing something obvious. I think what's happening is that the determined SCM type of PWD is "winning" as compared to the CODEMGR_WS specification in the (wx-ish) file list. I think (but am not sure) that the paths specified in the file list should take precedence in that case? (I'll file a bug, if so). -- Rich
