On 4/19/23 16:04, Simon Marchi wrote: > On 4/18/23 08:43, Tom de Vries wrote: >> On 4/5/23 11:01, Tom de Vries via Gdb-patches wrote: >>> On 4/4/23 13:35, Andrew Burgess wrote: >>>> Tom de Vries via Gdb-patches writes: >>>> >>>>> Add script gdb/contrib/make-check-all.sh, that's intended to function as a >>>>> drop-in replacement of make check, but excercising all host/target boards in >>>>> gdb/testsuite/boards. >>>>> >>>>> Shell-checked and tested on x86_64-linux. >>>> >>>> Hi Tom, >>>> >>>> Thanks for putting this together, I think this could be really useful. >>>> >>> >>> That's great to hear :) >>> >>>> I'm not a fan of the way you've split the patch description into email >>>> 0/1 and not included it with this commit. I think there's lots of >>>> useful information in there, and I'd much rather have the whole >>>> description included in the commit message -- it's much easier to find >>>> then rather than having to hunt on the mailing list in the future. >>>> >>> >>> Ack, I've reworked this into a patch rather than patch series. >>> >>>>> --- >>>>> gdb/contrib/make-check-all.sh | 255 ++++++++++++++++++++++++++++++++++ >>>>> 1 file changed, 255 insertions(+) >>>>> create mode 100755 gdb/contrib/make-check-all.sh >>>>> >>>>> diff --git a/gdb/contrib/make-check-all.sh b/gdb/contrib/make-check-all.sh >>>>> new file mode 100755 >>>>> index 00000000000..1befe418b81 >>>>> --- /dev/null >>>>> +++ b/gdb/contrib/make-check-all.sh >>>>> @@ -0,0 +1,255 @@ >>>>> +#!/bin/bash >>>>> + >>>>> +# Copyright (C) 2023 Free Software Foundation, Inc. >>>>> +# This program is free software; you can redistribute it and/or modify >>>>> +# it under the terms of the GNU General Public License as published by >>>>> +# the Free Software Foundation; either version 3 of the License, or >>>>> +# (at your option) any later version. >>>>> +# >>>>> +# This program is distributed in the hope that it will be useful, >>>>> +# but WITHOUT ANY WARRANTY; without even the implied warranty of >>>>> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the >>>>> +# GNU General Public License for more details. >>>>> +# >>>>> +# You should have received a copy of the GNU General Public License >>>>> +# along with this program. If not, see . >>>>> + >>>>> +# Run make check with all boards from gdb/testsuite/boards. >>>> >>>> I think you need some kind of usage test here too. Imagine a new >>>> developer finds this file and looks inside. Ideally, I think they >>>> should know how to use it, and what it will do for them without having >>>> to read any of the actual script content. >>>> >>> >>> Added. >>> >>>> On implementation, I wonder if it would be useful to provide a mechanism >>>> by which the gdb.sum and gdb.log files for each test run could be >>>> preserved? >>>> >>> >>> Done, I've renamed --keep to --keep-tmp, and added this new functionality under --keep. > > Maybe call it "keep-results"? Otherwise it's not clear what --keep keeps. > Done. >>> >>> I've also added "set -e" to make sure ^C aborts the script rather than just one make check invocation, and moved the tmpdir removal to a cleanup function that's executed on exit, to make sure ^C doesn't leave tmpdirs behind. >>> >> >> I'd like to commit this at the end of the week, if there are no further review comments. >> >> Simon, I remember discussing something similar with you a few years ago ( https://sourceware.org/pipermail/gdb-patches/2020-May/169003.html ), so I wondered if you have any comments on this approach. > > I don't recall this discussion at all :). But I like the idea of your > script, it lets people run tests on all target boards without having to > specifically know about each board. I never really test on boards other > than unix, native-gdbserver and native-extended-gdbserver, but I might > now with your script (probably not whole testsuite runs, but at least > when working on a test). > > When applying: > > Applying: Add make-check-all.sh > .git/rebase-apply/patch:232: trailing whitespace. > else > warning: 1 line adds whitespace errors. Ack, fixed. TIL: I can reproduce this using git show --check. > Would it make more sense to have this file under gdb/testsuite? Fine by me. [ FWIW, I usually look for scripts in a contrib dir, so my idea for the most logical place for it would be gdb/testsuite/contrib. If we'd add such a dir, we'd probably move the gdb/testsuite/*.py scripts there. Also, some from gdb/contrib, but I suppose it would leave gdb/contrib rather empty, that is only words.sh, ari/ and gdb-add-index.sh left AFAICT. ] Moved to gdb/testsuite. > I don't understand the comment about using --host-user and > --target-user. Are you suggesting that I create new users on my machine > (e.g. gdb-test-host and gdb-test-target), so that running with the > remote host and remote target boards pollute those users' home > directories instead of mine? To demonstrate the problem of polluting the home dir, after running gdb.base/advance.exp without --host-user and --target-user, I have these additional files in the home dir: ... advance0.o advance.c ccopts.c compiler.c ... And if I just use --target-user but not the --host-user one, we get the advance executable in addition. Yes, I do suggest creating new users. There are several reasons: - it prevents polluting your home dir, as some boards do, which on the flip side makes it easier to clean up if something was left, and make it easier to identify what was left. It also makes it easier to clean up inbetween boards, to prevent files copied by one board being used by the next board. - it prevents clashes in your home dir (I had a dir ~/dwz, which clashes with the exec from the gdb.dwarf2/dwz.exp test-case) - it's a good simulation of an actual remote host/target setup: - files from build can be made inaccessible from target/host and vice versa - pids from one user cannot be killed by other users, so errors like getting a pid on build and killing it on target/host will show up So, running the script with --host-user and --target-user gives you better checking, but for unsuspecting first-time users, the polluting home dir part will be the biggest thing I imagine, hence the comment. You say you don't understand the comment, but to me it seems you do so I'm not sure I understand what your concern is. Should the comment be improved, or do you want a change in functionality? FWIW, I've also written a script to add such test users, which I might submit after this gets committed, with the idea to both facilitate creation and standardize the result. > I also don't understand those `if true` conditions in main. Yeah, it was in preparation for an eventual ./make-check-all.sh --remote-target etc, but currently there's no need. Removed. > Finally, I'd appreciate a few comments here and there in the script, to > make further modifications easier. But other than that, it looks good > to me, it can't hurt to have that in the tree. I've tried adding a bit more comments, updated version attached. Thanks, - Tom