[Gluster-devel] NetBSD regression recovered (with patches)

[Emmanuel Dreyfus]

Hi

We now have the patches to recover NetBSD regression. This needs to be
reviewed and merged (it would be nice if it could be before release-3.7
branching):

http://review.gluster.org/10030
http://review.gluster.org/10032
http://review.gluster.org/10033
http://review.gluster.org/10034
http://review.gluster.org/9831
http://review.gluster.org/9944

I also have the following pending change that may deserve a look before
branching:
http://review.gluster.org/10017

Once all that will be merged, we will still have a few failing cases for
which I need some help (see details below):
tests/basic/afr/split-brain-resolution.t 
tests/basic/ec/
tests/basic/tier/tier.t
tests/encryption/crypt.t
tests/features/trash.t

I will disable them so that NetBSD regression vote can be useful. But it
would be nice if people coule help fixing them. Here are the details:

1) tests/basic/afr/split-brain-resolution.t 
100% reliable failure, I posted an analysis of it in 
<1m21hnh.yiu9vbyqrhg9M%manu at netbsd.org>


2) tests/basic/ec  tests have rare sprious failures, Xavier Hernandez
said he will look at it when time will allow it.


3) tests/basic/tier/tier.t has 100% reliable failure on nbslave70

[19:19:07] ./tests/basic/tier/tier.t .. 20/32
 md5: /d/backends/patchy1/d1/data2.txt: No such file or directory
[19:19:07] ./tests/basic/tier/tier.t .. 23/32 
not ok 23 Got "1" instead of "0"
md5: /d/backends/patchy1/d1/data3.txt: No such file or directory
[19:19:07] ./tests/basic/tier/tier.t .. 24/32 
not ok 24 Got "1" instead of "0"
umount: /mnt/glusterfs/0: Device busy
[19:19:07] ./tests/basic/tier/tier.t .. Failed 2/32 subtests 

./tests/basic/tier/tier.t (Wstat: 0 Tests: 32 Failed: 2)
  Failed tests:  23-24


Note that tier.t needs a portability patch not yet submitted:
diff --git a/tests/basic/tier/tier.t b/tests/basic/tier/tier.t
index 383d470..9cc754a 100755
--- a/tests/basic/tier/tier.t
+++ b/tests/basic/tier/tier.t
@@ -96,7 +96,7 @@ sleep 12
 uuidgen >> d1/data2.txt
 
 # Check promotion on read to slow tier
-echo 3 > /proc/sys/vm/drop_caches
+( cd $M0 && umount $M0 ) 
 cat d1/data3.txt
 sleep 5
 EXPECT_WITHIN $PROMOTE_TIMEOUT "0" file_on_fast_tier d1/data2.txt



4) tests/encryption/crypt.t has 100% reliable error on nbslave70. This
is annoying because that one passed before:

[19:21:21] ./tests/encryption/crypt.t .. 19/39 
ln: /mnt/glusterfs/0/testfile: Protocol error
not ok 20 
mv: rename /mnt/glusterfs/0/testfile to
/mnt/glusterfs/0/testfile-renamed: Protocol error
not ok 21 
diff: /mnt/glusterfs/0/testfile-symlink: No such file or directory
not ok 26 
[19:21:21] ./tests/encryption/crypt.t .. Failed 3/39 subtests 

./tests/encryption/crypt.t (Wstat: 0 Tests: 39 Failed: 3)
  Failed tests:  20-21, 26


5) tests/features/trash.t is completely broken. Note that
http://review.gluster.org/10033 patches trash.t for NetBSD compatibility

[19:24:33] ./tests/features/trash.t .. 1/65 No volumes present
[19:24:33] ./tests/features/trash.t .. 15/65 
not ok 15 
[19:24:33] ./tests/features/trash.t .. 18/65 
not ok 18 
[19:24:33] ./tests/features/trash.t .. 30/65 
not ok 30 
(...)
./tests/features/trash.t (Wstat: 0 Tests: 65 Failed: 16)
  Failed tests:  15, 18, 30, 33-34, 38-39, 57-65


-- 
Emmanuel Dreyfus
http://hcpnet.free.fr/pubz
manu at netbsd.org