Last modified: 2011-05-05 18:54:46 UTC
Type a unknown page name into the big box on Special:Export. Instead of indicating any error, one gets a file as if all is well. But if you examine it there is no payload at the bottom. Whats worse, if only some of the page names you give are unknown, well, the fact that they weren't found will also be unknown to you. Maybe add a box, checked, "ignore missing pages". Which if the user unchecks it, will bomb out if any page is not found. Or -- no box to check, just bomb out before returning any file. At least add a statement about what happens if a page is not found there on the boilerplate. Also please move "enter the titles in the text box below, one title per line" right next to the box.
*** Bug 28833 has been marked as a duplicate of this bug. ***
(In reply to comment #1) > *** Bug 28833 has been marked as a duplicate of this bug. *** This bug notes that, if the user specifies only missing pages (or URLs), Special:Export returns an XML file with no pages, just a siteinfo element. According to the reporter, Special:Import barfed on this (this is probably in MW 1.16 so it might've been fixed with the import rewrite). The exporter should never output anything that the importer can't import, so either of those two should be fixed.
Dupe bug included some indications that Special:Import also fails in an unpleasant way when given an import file with no pages; that should also be handled better if it hasn't already been. Note that Special:Export works by being given a set of zero or more titles, possibly expanding them into a larger list by fetching category listings, template usage, etc, and then goes through that list exporting whichever pages actually exist. If you ask for 5000 titles and only 4998 pages exist, it's correct behavior to receive an XML output file with 4998 pages in it. If you only asked for one, and it's not there, then yes that's information you may want to know. :) But keep in mind that some folks might even be deliberately asking for nonexistent pages just to get the export's siteinfo section -- this should not break. What's needed presumably is a way to report that the output file doesn't/won't contain specific titles that you asked for (whereas say a template invocation where the template doesn't exist presumably shouldn't warn you, since you didn't explicitly ask for the template but it was brought in from another page you did ask for which did exist). I guess this could be done by doing a separate hit to check all the pages you asked for before (or after? or ?) the actual export file build.
(In reply to comment #3) > some folks might even be deliberately asking for nonexistent pages just to get the export's siteinfo section -- this should not break. OK but if they are so smart they could use the API. Anyway, it's called "Export".