On January 17, 2022 1:42 PM, Junio C Hamano wrote: > <rsbecker@xxxxxxxxxxxxx> writes: > > > On January 14, 2022 9:19 PM, Junio wrote: > >> A release candidate Git v2.35.0-rc1 is now available for testing at the usual > places. > >> It is comprised of 467 non-merge commits since v2.34.0, contributed > >> by 80 people, 32 of which are new faces [*]. > > > > FYI: rc1 built/test (mostly) passes on both NonStop x86 and ia64 platforms. > > > > We seem to have a recurrence of a transient failure in t5562 from a few > releases ago. Retesting makes the problem disappear. > > > > Subtest 8 > > # > > # test_env HTTP_CONTENT_ENCODING="gzip" test_http_env upload > empty_body && > > # ! verify_http_result "200 OK" > > # > > Subtest 14 > > # > > # test_env HTTP_CONTENT_ENCODING="gzip" test_http_env > receive empty_body && > > # ! verify_http_result "200 OK" > > # > > > > Am I mistaken that using the ! verify_http_result construct is not recommended > or am I misremembering the discussion? > > > > Regards, > > --Randall > > I do not recall anything about that shell function. > > I found a half-thread with similar-looking report on an obscure system, since the > 2.25 era: > > https://lore.kernel.org/git/nycvar.QRO.7.76.6.2003252000200.46@tvgsbejvaqbjf . > bet/ > > And then from you that is even older, in the 2.21 era: > > https://lore.kernel.org/git/002201d4c560$5a319c40$0e94d4c0$@nexbridge.co > m/ > > But neither seems to lead to a discussion on the use of verify_http_result helper. We transitioned from the platform default ksh to bash for build/test of git right after that point that cleared the issue. It looks like there could be a regression in bash introduced on the platform very recently (June 2021) that may be responsible unless /bin/sh is referenced directly in the scripts. I will check. --Randall
