[03:59:47] AaronSchulz: so the disabling of DEFER_CONTENT_LENGTH etc regarding MediaWiki.php - is that effectively what <= 1.34 did? [04:00:16] My reading of the commit that introduced the issue wasn't that it introduced this use case, but that it altered it. but maybe the structure in place was effectively a no-op until then. [04:00:53] Its fine if it's not exactly like 1.34. I'm more asking to make sure I understand what the code did than anything else. [04:01:15] (and thus what the change is effectively) [04:01:32] RE: https://gerrit.wikimedia.org/r/c/mediawiki/core/+/672514 - [04:01:49] I'll rebase this on the doc/cleanup changes once you finish the review, as I might need to rebase twice otherwise. [04:15:07] Krinkle: the constants were added in 4f11b614544be8cb6198fbbef36e90206ed311bf in 1.35; I just removed the main part of that that mattered (headers) [04:15:50] Krinkle: which cleanups? [04:18:54] AaronSchulz: ah ok, but there were various similar conditions reached already. basically I'm asking: as far as you know, for all possible web server and php configutions, would you expect (or have intended) there to be any behavioural difference with regards to output headers and postsend defered/job running between 1.34 and 1.35 + this patch. [04:21:28] I'm aware we do still have improvement to convert more updates to jobs where possible by default, and ignore abort always at this piont. [04:21:55] I also see the added if ob_get_status() + end flush + flush still applies, which seems harmless indeed. [04:24:25] right [05:03:05] The error page handling has similar assumptions introduced but not so critical. Unlikely to become an issue by itself I suppose, we can sort that out afterward for the next release if the main approach ends up being different