Wip wjw freebsd compile by wjwithagen · Pull Request #7515 · ceph/ceph

wjwithagen · 2016-02-03T23:19:58Z

Enough changes to get ceph to compile on FreeBSD.

On FreeBSD the code now halts after testing runc-cli-tests
The next step will be to get the tests running....

tchaikov · 2016-02-04T16:15:35Z

src/common/ceph_time.cc

+        // documentation.
+#warning Falling back to CLOCK_REALTIME, may be slow.
+        clock_gettime(CLOCK_REALTIME, &ts);
+#endif


@wjwithagen FYI, #7466 by @adamemerson also addresses this issue.

On 4-2-2016 17:15, Kefu Chai wrote:

In src/common/ceph_time.cc
#7515 (comment):

@@ -49,7 +49,18 @@ namespace ceph {
coarse_real_clock::time_point coarse_real_clock::now(
const CephContext* cct) noexcept {
struct timespec ts;

clock_gettime(CLOCK_REALTIME_COARSE, &ts);
+#if defined(CLOCK_REALTIME_COARSE)

// Linux systems have _COARSE clocks.

clock_gettime(CLOCK_REALTIME_COARSE, &ts);
+#elif defined(CLOCK_REALTIME_FAST)

// BSD systems have _FAST clocks.

clock_gettime(CLOCK_REALTIME_FAST, &ts);
+#else

// And if we find neither, you may wish to consult your system's

// documentation.
+#warning Falling back to CLOCK_REALTIME, may be slow.

clock_gettime(CLOCK_REALTIME, &ts);
+#endif

@wjwithagen https://github.com/wjwithagen FYI, #7466
#7466 by @adamemerson
https://github.com/adamemerson also addresses this issue.

I know, but I "need" it to actually make any progress in getting things
compiling....
If there is an other way of doing this, just let me know...

This is a thing to consider anyways since quite a few people are
creating fixes for FreeBSD compiling. And several of the outstanding
pulls are conflicting and/or fixing the smae problem

And I'm not sure on how this will be handled when importing any of the
pulls. I sort of expect to rebase and fix merges once any of the
FreeBSD/Darwin/other pulls gets accepted.

--WjW

wjwithagen · 2016-02-05T08:59:34Z

@liewegas
I've thought a little bit more about it, and For the time being I'll require that FreeBSD build needs a softlinked /bin/bash to actually get things done.
And comment as such in a README.FreeBSD.
Just as a side note: I ran the script thru /usr/local/bin/dash on FreeBSD, and that just worked fine.

wjwithagen · 2016-02-05T10:58:12Z

'mmm

FAIL: test/mon/mon-created-time.sh

../test-driver: line 95: ./test/mon/mon-created-time.sh: Permission denied

Lets see if a rebase will help

tchaikov · 2016-02-05T18:00:14Z

mon-created-time.sh: Permission denied

it's fixed just now.

wjwithagen · 2016-02-05T18:45:13Z

@tchaikov
Thanx, that is going to clash with the chmod I just made in my tree.
But it'll work out.

wjwithagen · 2016-02-05T21:06:04Z

@tchaikov
I do not really understand why this errors
https://jenkins.ceph.com/job/ceph-pull-requests/1156/console

As far as I can tell, I'm up with HEAD. So that should not be the cause.

tchaikov · 2016-02-06T14:32:49Z

@wjwithagen this error does look suspicious.

@dachary i guess it might be a bug. i posted a pull request to address it at #7549.

wjwithagen · 2016-02-06T20:51:12Z

@tchaikov @dachary
I have not worked on any of the python files, and I try to keep my changes from actually modifying anything that changes the code for the Linux versions

wjwithagen · 2016-02-06T21:07:55Z

@tchaikov
I guess a rebase is in order.

tchaikov · 2016-02-08T05:57:37Z

@wjwithagen sorry for the confusion, i meant it was very likely a bug. and there are two other issues in testing got fixed recently. you might want to rebase against master to pick up them.

BTW, you might need to drop 82cda01 from your pull request.

wjwithagen · 2016-02-08T12:52:33Z

@tchaikov
Before I mess up my tree again:
How do I "drop" your commit 82cda01??
I could revert, but that will give a commit message

wjwithagen · 2016-02-08T13:31:26Z

@tchaikov
For FreeBSD boost again seems to work. Thanx.

liewegas · 2016-02-08T13:51:56Z

git rebase -i 82cda01^
and then delete the first line (the one with that commit)

On Mon, 8 Feb 2016, Willem Jan Withagen wrote:

@tchaikov
Before I mess up my tree again:
How do I "drop" your commit 82cda01??
I could revert, but that will give a commit message

—
Reply to this email directly or view it on GitHub.[AAabh4peTIQJsjucdntKb6HlaUHdgxfGks5piIctgaJpZM4HS7vv.gif]

wjwithagen · 2016-02-08T14:20:40Z

Some more errors which I cannot place:
PASS: test/mon/osd-pool-create.sh
PASS: test/mon/mon-ping.sh
PASS: test/mon/mon-created-time.sh
/tmp/hudson8518538414639280563.sh: line 2: 2508 Killed timeout 7200 ./run-make-check.sh
Build step 'Execute shell' marked build as failure

I guess that a watchdog kills things after 2 hours of running?
But I have no idea where the script actually gets stuck

I see that the script will not run on FreeBSD due to missing RBD, but I have not yet come to that.

liewegas · 2016-02-08T14:30:48Z

make check -j1 and ps xf might help narrow down which test gets hung...

On Mon, 8 Feb 2016, Willem Jan Withagen wrote:

Some more errors which I can place:
PASS: test/mon/osd-pool-create.sh
PASS: test/mon/mon-ping.sh
PASS: test/mon/mon-created-time.sh
/tmp/hudson8518538414639280563.sh: line 2: 2508 Killed timeout 7200 ./run-make-check.sh
Build step 'Execute shell' marked build as failure

I guess that a watchdog kills things after 2 hours of running?
But I have no idea where the script actually gets stuck

I see that the script will not run on FreeBSD due to missing RBD, but I have not yet come to that.

—
Reply to this email directly or view it on GitHub.[AAabhwyZlhI0s6g5WjE5b08HIApgG04oks5piJvEgaJpZM4HS7vv.gif]

wjwithagen · 2016-02-08T14:51:58Z

@liewegas
Eh, right but how do I do that for a Jenkins process.

I would expect ./run-make-check.sh to be the cullpit, but then again I cannot run that here since has RBD.

Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

But the code seems to compile even without it. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

We currently do not build Bluestore on FreeBSD due to incompatible AIO Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

…onstexpr Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

Clang refuses to promote an address to int64_t by itself Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

As long as FreeBSD has nog RBD device, all tests for RBD will not work. run-cli-tests needs to be passed before any of the other tests are build/run, so ATM we exclude the RBD tests and complete with success under FreeBSD Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

This is not available under FreeBSD, where the Posix variant -E is. Rewritten the tests to accept any whitespace type, instead of just TAB Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

…ockets.h> Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

…ndle MAP_ANONYMOUS The mapping is not backed by any file; its contents are initialized to zero. The fd and offset arguments are ignored; however, some implementations require fd to be -1 if MAP_ANONYMOUS (or MAP_ANON) is specified, and portable applications should ensure this. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

If certain properties of the disks are essential to test here, new tests need to be written specific for the FreeBSD situation. For now tests are run on a ZFS only system. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

* Could use 'df -T' for that, if really needed. * /proc/mounts is only available in linux emulation, and is found in /compat/linux * Also wrapped some long lines. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

… m are const. This is a GCC extention that is not available in Clang Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

This require mounting the linprocfs system: - kldload linprocfs - in /etc/fstab: linproc /compat/linux/proc linprocfs rw 0 0 Cannot mount in /proc, since that is taken bij FreeBSD itself Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD does not have PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP defined So we need to create the mutex ourselves. This is copied from older code in previous commits. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

stat -c '%T' does not give fstype Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

So do not use a hardcode struct to initialize. Instead assign all the required fields individually Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

_check_disk_write_cache is #ifdef to be for Linux only. FreeBSD does not have hdparm tool to check disk setting. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

This is hard to test from the program itself, and would need to be tested from a script/program that actually calls this test. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

liewegas added common build/ops needs-qa wip-sage-testing labels Feb 4, 2016

tchaikov reviewed Feb 4, 2016
View reviewed changes

liewegas removed needs-qa wip-sage-testing labels Feb 4, 2016

wjwithagen force-pushed the wip-wjw-freebsd-compile branch from 64342dc to fd23298 Compare February 5, 2016 15:13

wjwithagen force-pushed the wip-wjw-freebsd-compile branch 2 times, most recently from 6e8b98b to c2c5aad Compare February 5, 2016 20:55

wjwithagen force-pushed the wip-wjw-freebsd-compile branch 4 times, most recently from 52e303b to 382422a Compare February 11, 2016 10:34

wjwithagen force-pushed the wip-wjw-freebsd-compile branch 3 times, most recently from 3b31e53 to d91e126 Compare May 22, 2016 20:09

wjwithagen force-pushed the wip-wjw-freebsd-compile branch from d91e126 to 53d8781 Compare May 27, 2016 12:14

wjwithagen added 22 commits May 29, 2016 14:00

FreeBSD/test: ceph_test_objectstore can be build as well.

9bf453a

Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/pthread: The non-POSIX routines are in a different include file

b1a8e3e

Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/features: This file is not on FreebSD.

dac0394

But the code seems to compile even without it. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/Bluestore: exclude Bluestore types from being generated.

9e8a43c

We currently do not build Bluestore on FreeBSD due to incompatible AIO Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/xattr: Use a different include, and fix CONSTS

c8415ef

Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/malloc: Mallinfo is not available in FreeBSD

dfaceb2

Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

common/test_time.cc: Clang and GCC have different opinions on us of c…

90bd446

…onstexpr Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

common/Graylog.cc: GCC and Clang handle casting differently.

c55a14c

Clang refuses to promote an address to int64_t by itself Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/POSIX: tests in src/test/cli used grep -P

968c05e

This is not available under FreeBSD, where the Posix variant -E is. Rewritten the tests to accept any whitespace type, instead of just TAB Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/sockets: Socket constants and functions are defined in <sys/s…

e70bdf6

…ockets.h> Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/ceph-disk.sh: stat -c '%T' does not give fstype

8f2cfa7

* Could use 'df -T' for that, if really needed. * /proc/mounts is only available in linux emulation, and is found in /compat/linux * Also wrapped some long lines. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/Clang: buffer::ptr enc[k + m] declaration only works when k &…

7fbbc40

… m are const. This is a GCC extention that is not available in Clang Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/mutex_debug.h: Manually create the debug mutex

934a908

FreeBSD does not have PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP defined So we need to create the mutex ourselves. This is copied from older code in previous commits. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/ceph-helpers.sh test_pidfile.sh: fix more use of stat -C %T

14f04c4

stat -c '%T' does not give fstype Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/pidfile.cc: struct flock is different on FreeBSD.

2615936

So do not use a hardcode struct to initialize. Instead assign all the required fields individually Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD/FileJournal.cc NO verifing cache functionality on FreeBSD

1090776

_check_disk_write_cache is #ifdef to be for Linux only. FreeBSD does not have hdparm tool to check disk setting. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

FreeBSD//cross_process_sem.cc: anonymous mmap needs -1 as fd-handle

69c7d5a

Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

test/confutils.cc: Report that the atexit routine is really excuted.

2ab6bf8

This is hard to test from the program itself, and would need to be tested from a script/program that actually calls this test. Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>

wjwithagen force-pushed the wip-wjw-freebsd-compile branch from 53d8781 to 2ab6bf8 Compare May 29, 2016 12:00

wjwithagen closed this Jul 28, 2016

wjwithagen mentioned this pull request Jul 28, 2016

cmake: fix the build on AArch64 #10427

Closed

wjwithagen deleted the wip-wjw-freebsd-compile branch January 4, 2017 10:23

Conversation

wjwithagen commented Feb 3, 2016

Uh oh!

tchaikov Feb 4, 2016

Choose a reason for hiding this comment

Uh oh!

wjwithagen Feb 4, 2016

Choose a reason for hiding this comment

Uh oh!

wjwithagen commented Feb 5, 2016

Uh oh!

wjwithagen commented Feb 5, 2016

Uh oh!

tchaikov commented Feb 5, 2016

Uh oh!

wjwithagen commented Feb 5, 2016

Uh oh!

wjwithagen commented Feb 5, 2016

Uh oh!

tchaikov commented Feb 6, 2016

Uh oh!

wjwithagen commented Feb 6, 2016

Uh oh!

wjwithagen commented Feb 6, 2016

Uh oh!

tchaikov commented Feb 8, 2016

Uh oh!

wjwithagen commented Feb 8, 2016

Uh oh!

wjwithagen commented Feb 8, 2016

Uh oh!

liewegas commented Feb 8, 2016

Uh oh!

wjwithagen commented Feb 8, 2016

Uh oh!

liewegas commented Feb 8, 2016

Uh oh!

wjwithagen commented Feb 8, 2016

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants