Code coverage gone down after 8.2.0377

[Yegappan Lakshmanan]

Hi all,

I see that the code coverage has gone down from 84.55% to 84.29%

after patch 8.2.0377 (no CI test for a big-endian system). I am not sure

how enabling additional builds reduced the code coverage.

- Yegappan

[Bram Moolenaar]

Indeed. The "tracked lines" did not change, the "Lines hit" did go
down. No idea how adding another configuration can cause that.

You could perhaps dig deeper into what "lines hit" disappeared:
before:
https://codecov.io/gh/vim/vim/tree/8210693795d6d0d51bf5b70674d4539cdde0330b
after:
https://codecov.io/gh/vim/vim/tree/d47e6f0b4cc82e3ccdc4605bb1811861a5b2d115

--
Place mark here ->[ ]<- if you want a dirty monitor.

/// Bram Moolenaar -- Br...@Moolenaar.net -- http://www.Moolenaar.net \\\
/// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ \\\
\\\ an exciting new programming language -- http://www.Zimbu.org ///
\\\ help me help AIDS victims -- http://ICCF-Holland.org ///

[Dominique Pellé]

Bram Moolenaar <Br...@moolenaar.net> wrote:

> Yegappan wrote:
>
> > I see that the code coverage has gone down from 84.55% to 84.29%
> > after patch 8.2.0377 (no CI test for a big-endian system). I am not sure
> > how enabling additional builds reduced the code coverage.
>
> Indeed. The "tracked lines" did not change, the "Lines hit" did go
> down. No idea how adding another configuration can cause that.
>
> You could perhaps dig deeper into what "lines hit" disappeared:
> before:
> https://codecov.io/gh/vim/vim/tree/8210693795d6d0d51bf5b70674d4539cdde0330b
> after:
> https://codecov.io/gh/vim/vim/tree/d47e6f0b4cc82e3ccdc4605bb1811861a5b2d115

It seems to be mostly GUI and X clipboard code that
is now less covered.

In the diff of patch 8.2.0377 I see:

```
@@ -73,9 +73,13 @@ anchors:
fi
before_script:
# Start virtual framebuffer to be able to test the GUI. Does
not work on OS X.
- - export DISPLAY=:99.0
- - sh -e /etc/init.d/xvfb start && sleep 3
- - sudo modprobe snd-dummy
+ - |
+ if [[ "${TEST}" =~ gui ]]; then
+ export DISPLAY=:99.0
+ sh -e /etc/init.d/xvfb start && sleep 3
+ fi
+ - |
+ [ "${TRAVIS_CPU_ARCH}" = s390x ] || sudo modprobe snd-dummy
- sudo usermod -a -G audio $USER
- do_test() { sg audio "sg $(id -gn) '$*'"; }
```

It might be elated to that. Perhaps we don't enter
the 'if' branch for any platform?

Dominique

[Bram Moolenaar]

Looking at the output for "huge-testgui+coverage/gcc", both say "Skipped
44 Tests", so that looks OK.

--
hundred-and-one symptoms of being an internet addict:
259. When you enter your name in the AltaVista search engine, the top ten
matches do indeed refer to you.

[Yegappan Lakshmanan]

Hi,

I think, even in non-gui builds, some of the code path (e.g. clipboard)

depend on the DISPLAY being set.

I don't know whether this change is relevant to enabling the s390 builds.

This change should be reverted as this reduces the code coverage.

- Yegappan

[James McCoy]

On Sun, Mar 15, 2020, 12:19 Yegappan Lakshmanan <yega...@gmail.com> wrote:

I think, even in non-gui builds, some of the code path (e.g. clipboard)

depend on the DISPLAY being set.

I don't know whether this change is relevant to enabling the s390 builds.

This change should be reverted as this reduces the code coverage.

It was related to the s390x changes, as that was failing to run. I think I know a better way to address that. I'll work on a PR.

[Bram Moolenaar]

if_xcmdsrv.c test also has lower coveage.

> I don't know whether this change is relevant to enabling the s390 builds.
> This change should be reverted as this reduces the code coverage.

Can you make a pull request for putting back the DISPLAY setting?

--
hundred-and-one symptoms of being an internet addict:

271. You collect hilarious signatures from all 250 mailing lists you
are subscribed to.

[Yegappan Lakshmanan]

Hi Bram,

It looks like James McCoy is working on a PR to properly address this

without breaking the s390 builds.

Regards,

Yegappan