[testing] How do you interpret the SHA1 revision (outdated results)?

older
Compile time reflection for IPC...

John Maddock

23 Jan 2014 23 Jan '14

10:14 a.m.

So... I'm looking into Boost.Regex and I'm seeing some failures which appear to refer to outdated sources (line numbers are all wrong for current source), for example: http://www.boost.org/development/tests/develop/developer/output/marshall-mac... But how on earth do I check this? Reported run date is yesterday, but this was fixed/changed on Dec 20th: almost a month ago. And the revision listed on the page is a SHA1 which I can't seem to relate to anything useful? Thanks, John.

Show replies by date

Vladimir Prus

23 Jan 23 Jan

10:35 a.m.

New subject: [testing] How do you interpret the SHA1 revision (outdated results)?

On 23.01.2014 14:14, John Maddock wrote:

...

So... I'm looking into Boost.Regex and I'm seeing some failures which appear to refer to outdated sources (line numbers are all wrong for current source), for example: http://www.boost.org/development/tests/develop/developer/output/marshall-mac...

But how on earth do I check this? Reported run date is yesterday, but this was fixed/changed on Dec 20th: almost a month ago. And the revision listed on the page is a SHA1 which I can't seem to relate to anything useful?

This SHA1 is superproject one: commit 3d21a16db5ada89d4c5875074a87ba81239048a0 Author: Automated Commit Date: Mon Jan 20 04:30:03 2014 -0500 Update phoenix. And it's on master branch. Here's the corresponding regex revision: $ git ls-tree 3d21a16db5ada89d4c5875074a87ba81239048a0:libs | grep regex 160000 commit a3075dd98ffa4ffea8cf6331e1ef19dc1b42459e regex That revision is the tip of the master branch, and is indeed, from Dec 7. - Volodya

John Maddock

11:18 a.m.

New subject: [testing] How do you interpret the SHA1 revision(outdated results)?

...

This SHA1 is superproject one:

commit 3d21a16db5ada89d4c5875074a87ba81239048a0 Author: Automated Commit Date: Mon Jan 20 04:30:03 2014 -0500

Update phoenix.

And it's on master branch. Here's the corresponding regex revision:

$ git ls-tree 3d21a16db5ada89d4c5875074a87ba81239048a0:libs | grep regex 160000 commit a3075dd98ffa4ffea8cf6331e1ef19dc1b42459e regex

That revision is the tip of the master branch, and is indeed, from Dec 7.

Great. Two questions: 1) How come the Mac tests are being run on 2 month old sources? 2) How did you figure that out? If I run the git ls-tree command above on my working system I get a "not a valid object name" error, plus not everyone has a Unix command line to hand... Thanks, John.

Vladimir Prus

11:50 a.m.

New subject: [testing] How do you interpret the SHA1 revision(outdated results)?

On 23.01.2014 15:18, John Maddock wrote:

...

...
This SHA1 is superproject one:

commit 3d21a16db5ada89d4c5875074a87ba81239048a0 Author: Automated Commit Date: Mon Jan 20 04:30:03 2014 -0500

Update phoenix.

And it's on master branch. Here's the corresponding regex revision:

$ git ls-tree 3d21a16db5ada89d4c5875074a87ba81239048a0:libs | grep regex 160000 commit a3075dd98ffa4ffea8cf6331e1ef19dc1b42459e regex

That revision is the tip of the master branch, and is indeed, from Dec 7.

Great. Two questions:

1) How come the Mac tests are being run on 2 month old sources?

Sorry for being vague above. They are run on master branch of the super project, which points to master branch of regex, which is that old. I am unsure what's the value of testing master branch of superproject at this point. It's too early.

...

2) How did you figure that out?

If I run the git ls-tree command above on my working system I get a "not a valid object name" error, plus not everyone has a Unix command line to hand...

Did you do "git fetch" in the superproject? As for Unix command line, the approach that Michael used will work as well - look at the tree at github and it should have subproject references, like so: - Go to https://github.com/boostorg/boost/commit/3d21a16db5ada89d4c5875074a87ba81239... - Click 'browse code' - Go to 'libs' and record the short id of regex. - Manually paste that id into URL, like so: https://github.com/boostorg/regex/commit/a3075dd Yes, it's ugly roundabout way. - Volodya

Daniel James

11:57 a.m.

New subject: [testing] How do you interpret the SHA1 revision(outdated results)?

On 23 January 2014 11:50, Vladimir Prus wrote:

...

Sorry for being vague above. They are run on master branch of the super project, which points to master branch of regex, which is that old. I am unsure what's the value of testing master branch of superproject at this point. It's too early.

It's a mistake since it's part of the develop report, I don't know if it's a problem in the regression scripts, or user error.

...

...
2) How did you figure that out?

If I run the git ls-tree command above on my working system I get a "not a valid object name" error, plus not everyone has a Unix command line to hand...

Did you do "git fetch" in the superproject? As for Unix command line, the approach that Michael used will work as well - look at the tree at github and it should have subproject references, like so:

- Go to https://github.com/boostorg/boost/commit/3d21a16db5ada89d4c5875074a87ba81239... - Click 'browse code' - Go to 'libs' and record the short id of regex. - Manually paste that id into URL, like so: https://github.com/boostorg/regex/commit/a3075dd

Yes, it's ugly roundabout way.

There's another thread about improving the information in the report, which could include a link to github. But apparently I have to implement it, which might not happen soon. I could possibly improve the commit messages. An easy change would be to include the submodule's branch (not the super-project's since commits can appear on multiple branches), I might try doing that tonight.

Daniel James

10:14 p.m.

New subject: [testing] How do you interpret the SHA1 revision(outdated results)?

On 23 January 2014 11:57, Daniel James wrote:

...

I could possibly improve the commit messages. An easy change would be to include the submodule's branch (not the super-project's since commits can appear on multiple branches), I might try doing that tonight.

OK, I've done that. You can see the new message at: https://github.com/boostorg/boost/commit/f099295d881f8ec551fe9f56086f50a6648... Let me know if you have any ideas or opinions. I want to include more information in the commit (maybe the new submodule commits or maybe the dates from the updated submodules), but I don't currently have a good way to get information from the submodules themselves, so I can't do much at the moment.

Beman Dawes

24 Jan 24 Jan

12:40 p.m.

New subject: [testing] How do you interpret the SHA1 revision(outdated results)?

On Thu, Jan 23, 2014 at 5:14 PM, Daniel James wrote:

...

On 23 January 2014 11:57, Daniel James wrote:

...
I could possibly improve the commit messages. An easy change would be to include the submodule's branch (not the super-project's since commits can appear on multiple branches), I might try doing that tonight.

OK, I've done that. You can see the new message at:

https://github.com/boostorg/boost/commit/f099295d881f8ec551fe9f56086f50a6648...

Nice! If you want to shorten the SHA to the first six digits, that's plenty and shortening is a commonplace in the git world.

...

Let me know if you have any ideas or opinions. I want to include more information in the commit (maybe the new submodule commits or maybe the dates from the updated submodules), but I don't currently have a good way to get information from the submodules themselves, so I can't do much at the moment.

Timestamp? I didn't follow the discussion of timestamps closely, so am not sure if a timestamp could be generated that would be useful, however. --Beman

Daniel James

23 Jan 23 Jan

12:02 p.m.

New subject: [testing] How do you interpret the SHA1 revision(outdated results)?

On 23 January 2014 11:18, John Maddock wrote:

...

If I run the git ls-tree command above on my working system I get a "not a valid object name" error, plus not everyone has a Unix command line to hand...

I hope most graphical tools have somewhere to enter a commit hash.

Beman Dawes

9:44 p.m.

New subject: [testing] How do you interpret the SHA1 revision(outdated results)?

On Thu, Jan 23, 2014 at 7:02 AM, Daniel James wrote:

...

On 23 January 2014 11:18, John Maddock wrote:

...
If I run the git ls-tree command above on my working system I get a "not a valid object name" error, plus not everyone has a Unix command line to hand...

I hope most graphical tools have somewhere to enter a commit hash.

TortoiseGit: Select TortoiseGit from the Windows Explorer right-click drop-down menu for the Boost tree directory, select "Show log" from the TortoiseGit drop-down, and copy the SHA-1 (or part of the SHA-a) into the Filter box (with the little magnifying glass icon). If you get only a single "Working dir changes" entry with a SHA of all zeros, it means the SHA you entered wasn't found. --Beman

John Maddock

24 Jan 24 Jan

9:35 a.m.

New subject: [testing] How do you interpret the SHA1revision(outdated results)?

...

TortoiseGit:

Select TortoiseGit from the Windows Explorer right-click drop-down menu for the Boost tree directory, select "Show log" from the TortoiseGit drop-down, and copy the SHA-1 (or part of the SHA-a) into the Filter box (with the little magnifying glass icon). If you get only a single "Working dir changes" entry with a SHA of all zeros, it means the SHA you entered wasn't found.

That doesn't really help: all it shows you is the automated commit in the super project that that hash refers to, it tells you nothing about which version the particular sub-libraries are on. Ah wait... right click and "Browse repository" gives you a view of the repro * at that commit*, you can then drill down to the library(s) of interest and right click and select "Show log" to get the information you need. OK, that's almost useable. Now we just need to sort out why Marshall's results run on the master branch are showing up for develop? Thanks, John.

Beman Dawes

12:45 p.m.

New subject: [testing] How do you interpret the SHA1revision(outdated results)?

On Fri, Jan 24, 2014 at 4:35 AM, John Maddock wrote:

...

...
... Now we just need to sort out why Marshall's results run on the master branch are showing up for develop?

Yes, that is a serious worry. Is this happening for all test runners or just Marshall? --Beman