[Freemint-list] Git for FreeMiNT (was Re: [MiNT] fVDI in FreeMiNT CVS?)

[Miro Kropáček]

Small heads up...

I've got not so encouraging news, we're drowned in regressions. On FreeMiNT
1.18 (the official one), bash 4.2 and perl 5.18 I'm able to get the best
results:

>From *12913* tests excluding known breakages and external tools' tests is
*12501* 'ok' and *416* 'not ok'. That doesn't sound that bad, does it. The
failing ones are for instance umask problems, stdin streaming, handling of
backslashes, escaping, locale tests, ... there's about 4-5 groups of
failing git tests (formatting, diff tests, work tree, submodules,
patch/email ...) which doesn't seem awfully hard to fix.

What concerns me is the fact that when I switch to bash 4.3 (from Vincent's
site), the test environment stops working (the very basic tests are
failing) and when I switch to the current 1.19, some tests run endlessly.
And since 'git' and its test suite isn't one exe but like one zillion of
them, it's super hard to track down the whole flow happening within a test.
But from what I saw, it's nothing trivial, for instance the 1.19 kernel
issue is somehow related to running multiple processes even though I hardly
can see a commit related to it in the 1.19 branch.

And yes, I was very careful to isolate the symptoms, I always tested the
new builds (bash, freemint, perl, coretutils, toswin2) on a clean image.
Ideally, if FreeMiNT + mintlib would be on github, we could setup a travis
ci client which would run an aranym image with test suites from perl,
openssl, git, gcc ... (any big project with nice test coverage) after each
commit so we could detect side effects very quickly.


On 7 November 2016 at 10:02, Miro Kropáček <miro.kropacek at gmail.com> wrote:

> then it also reveals a hardcore crash in ext2.xfs -- I can take down whole
>> Aranym (even in the MMU mode!) if I run the tests from an ext2 partition
>> (works fine on aranym's hostfs, irony, isn't it).
>>
> OK, another one hunted down, now the praise goes to Vincent and his
> recompiled coreutils.
>
> Debug output of the kernel revealed:
>
> pid 286 (rm): assert (`fs->fsflags & FS_EXT_3') failed at line 48 of
> ../xfs_xdd.c.
>
> what is a function in xfs_xdd.c:
>
> long
> getxattr(FILESYS *fs, fcookie *fc, XATTR *xattr)
> {
>         STAT stat;
>         long r;
>
>         assert(fs->fsflags & FS_EXT_3);
> ...
>
> so I installed a newer 'rm' from aforementioned package and voila. I think
> asserts like this should be more user-visible, there's no indication of an
> error unless you change the debug level...
>
> After this "fix", future looks much brighter -- except the failing umask
> test, there is a couple of tests failing related to so called filters, a
> few EOF/backslash related, ... so far I've been able to run a few hundreds
> tests and there's about 20 of them failing.
>
> The worst is that some bugs in stdin processing just stop the testing
> process and I have to manually disable that script, argh.
>
> --
> MiKRO / Mystic Bytes
> http://mikro.atari.org
>
> _______________________________________________
> Freemint-list mailing list
> Freemint-list at mail.atariforge.org
> http://mail.atariforge.org/mailman/listinfo/freemint-list
>
>


-- 
MiKRO / Mystic Bytes
http://mikro.atari.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.atariforge.org/pipermail/freemint-list/attachments/20161110/29971718/attachment.html