From: Han-Wen Nienhuys Date: Wed, 10 Jun 2020 06:56:43 GMT Subject: Re: [PATCH v16 00/14] Reftable support git-core Message-ID: In-Reply-To: On Wed, Jun 10, 2020 at 1:14 AM Junio C Hamano wrote: > > Some issues: > > > > * many tests inspect .git/{logs,heads}/ directly. > > * worktrees broken. > > Are these with or without GIT_TEST_REFTABLE=1? worktrees are broken with reftable, because HEAD is per-worktree ref, and I haven't decided how to handle those for reftable. > The patches (with the fix-up I saw today) applied on top of your > refs clean-up topic seems to pass the tests (without setting > GIT_TEST_REFTABLE) as the whole, but when merged at the tip of 'pu', > many tests fail (please see [*1*], for examples). I suspect that > there may be some topics in flight that add more direct access to > the .git/refs/ hierarchy, and/or there may be stupid mismerges (I'd > especially appreciate it if you can spot any). Curious; I'll have a look. There is also: reftable/stack_test.c:27:7: error: incompatible pointer types initializing 'PREC_DIR *' with an expression of type 'DIR *' [-Werror,-Wincompatible-pointer-types] DIR *dir = fdopendir(fd); on OSX. What is the proper dialect for reading out a directory within the git source code? opendir and fdopendir are POSIX, so I'm surprised this fails. -- Han-Wen Nienhuys - hanwenn@gmail.com - http://www.xs4all.nl/~hanwen