On 2022-05-07 20:52:41 [+0200], Paul Gevers wrote: > Hi Sebastian, Hi Paul,
> On 07-05-2022 18:22, Sebastian Andrzej Siewior wrote: > > Usertags: flaky > > Why do you conclude that? Normally we call something flaky if it has a > reasonable amount of failures in pure testing environments (so no migration > runs). I'm not seeing that for tunnel4 on amd64, nor arm64. Maybe I missunderstood. According to the tracking page, it failed everywhere: | autopkgtest for stunnel4/3:5.63-1: | amd64: Regression ♻ (reference ♻), | arm64: Regression ♻ (reference ♻), | armhf: Regression ♻ (reference ♻), | i386: Regression ♻ (reference ♻), | ppc64el: Regression ♻ (reference ♻), | s390x: Regression ♻ (reference ♻) Does this mean flaky or is this something in case it fails _always_ and not just randomly swi-prolog/8.4.2+dfsg-2 on i386 while it passed on other arches. > > that the error is that it was compiled against one version (1.1.1n) > > and then tested against another version (1.1.1o)? > > stunnel4 -version reports: > > | Compiled with OpenSSL 1.1.1n 15 Mar 2022 > > | Running with OpenSSL 1.1.1o 3 May 2022 > > > > and it is fine, the ABI is stable. > > If your claim is true (and I trust it is), I do agree that it seems that the > test (and I guess this comes from a runtime check) is too strict. Retitling > accordingly. Oki, thanks. > Paul Sebastian