Re: New test failures in fs/vfs/t_io

[Taylor R Campbell <riastradh%NetBSD.org@localhost>]

   Date: Wed, 1 Apr 2015 20:16:38 +0300
   From: Andreas Gustafsson <gson%gson.org@localhost>

   Earlier today, I wrote:
   > Fourteen test cases of the fs/vfs/t_io test started failing on the TNF
   > testbed during a recent period when the build and/or install were
   > broken, between 2015-03-28 16:13:38 and 2015-03-29 19:35:58 UTC.

   Never mind, I see now that these are not new failures of existing test
   cases, but new test cases that were added by riastradh during the time
   period in case.

Correct.  There are some new tests that are broken due to a bug in
rump, but the bug I added tests for is fixed in most file systems.
Sorry for the trouble and confusion.

What happened was:

1. I noticed a bug in ffs: read(fd, NULL, 1) on a file with
O_SYNC|O_RSYNC on a mount without noatime would return success when it
should return EFAULT.

2. I wrote some tests for this and found rump_sys_read(fd, NULL, 1)
crashes instead of returning EFAULT.

3. I committed the xfail tests and a workaround for the rump bug:

https://mail-index.netbsd.org/source-changes/2015/03/28/msg064369.html
https://mail-index.netbsd.org/source-changes/2015/03/28/msg064371.html

4. I fixed the ffs bug and removed the xfail:

https://mail-index.netbsd.org/source-changes/2015/03/28/msg064381.html

5. I noticed that my workaround broke the non-rump case (causing the
install breakage), committed a different workaround that would be
guaranteed not to affect non-rump, and then backed it out at pooka's
request:

https://mail-index.netbsd.org/source-changes/2015/03/29/msg064409.html
https://mail-index.netbsd.org/source-changes/2015/03/29/msg064410.html
https://mail-index.netbsd.org/source-changes/2015/03/29/msg064411.html

Fixing the rump bug properly remains to be done, so the tests are
broken for now.