There's no need to have different CI process between macOS and FreeBSD
as test suite has been fixed on macOS.

Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
---
 ci/cirrus/build.yml | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/ci/cirrus/build.yml b/ci/cirrus/build.yml
index 912284b906..dc1cccd135 100644
--- a/ci/cirrus/build.yml
+++ b/ci/cirrus/build.yml
@@ -20,5 +20,4 @@ build_task:
     - git reset --hard "$CI_COMMIT_SHA"
   build_script:
     - meson build --prefix=$(pwd)/install-root
-    - if test "$(uname)" = "FreeBSD"; then ninja -C build dist; fi
-    - if test "$(uname)" = "Darwin"; then ninja -C build && ninja -C build install; fi
+    - ninja -C build dist
-- 
2.29.2

On 11/8/20 10:24 PM, Roman Bolshakov wrote:
> There's no need to have different CI process between macOS and FreeBSD
> as test suite has been fixed on macOS.
> 
> Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
> ---
>   ci/cirrus/build.yml | 3 +--
>   1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/ci/cirrus/build.yml b/ci/cirrus/build.yml
> index 912284b906..dc1cccd135 100644
> --- a/ci/cirrus/build.yml
> +++ b/ci/cirrus/build.yml
> @@ -20,5 +20,4 @@ build_task:
>       - git reset --hard "$CI_COMMIT_SHA"
>     build_script:
>       - meson build --prefix=$(pwd)/install-root
> -    - if test "$(uname)" = "FreeBSD"; then ninja -C build dist; fi
> -    - if test "$(uname)" = "Darwin"; then ninja -C build && ninja -C build install; fi
> +    - ninja -C build dist
> 

If we go with --timeout-multiplier= as I'm suggesting in 3/4 then this 
can't be ninja, but meson. I'm not sure what the whole point of 'ninja' 
is at this point, sorry.

Michal

On Fri, 2020-11-13 at 16:58 +0100, Michal Privoznik wrote:
> On 11/8/20 10:24 PM, Roman Bolshakov wrote:
> > -    - if test "$(uname)" = "FreeBSD"; then ninja -C build dist; fi
> > -    - if test "$(uname)" = "Darwin"; then ninja -C build && ninja -C build install; fi
> > +    - ninja -C build dist
> 
> If we go with --timeout-multiplier= as I'm suggesting in 3/4 then this 
> can't be ninja, but meson. I'm not sure what the whole point of 'ninja' 
> is at this point, sorry.

ninja is the low-level tool, so in some cases (notably running the
test suite) having meson call ninja instead of invoking the latter
directly can enable additional features.

In this case, I assume 'ninja dist' will use 'ninja test' rather than
'meson test', so we might have to do something like

  - if test "$(uname)" = "FreeBSD"; then cd build && ninja dist; fi
  - if test "$(uname)" = "Darwin"; then cd build && ninja && meson test --timeout-multiplier=X && ninja install; fi

to run the test suite on macOS without hitting the timeout.

-- 
Andrea Bolognani / Red Hat / Virtualization

On 11/13/20 8:08 PM, Andrea Bolognani wrote:
> On Fri, 2020-11-13 at 16:58 +0100, Michal Privoznik wrote:
>> On 11/8/20 10:24 PM, Roman Bolshakov wrote:
>>> -    - if test "$(uname)" = "FreeBSD"; then ninja -C build dist; fi
>>> -    - if test "$(uname)" = "Darwin"; then ninja -C build && ninja -C build install; fi
>>> +    - ninja -C build dist
>>
>> If we go with --timeout-multiplier= as I'm suggesting in 3/4 then this
>> can't be ninja, but meson. I'm not sure what the whole point of 'ninja'
>> is at this point, sorry.
> 
> ninja is the low-level tool, so in some cases (notably running the
> test suite) having meson call ninja instead of invoking the latter
> directly can enable additional features.

And I guess that's my problem with it. The --timeout-multiplier= is 
argument of meson, but if I want to run a single threaded build (useful 
if I mess up something in a header file that's included from everywhere) 
there's no way to pass "-j1" to meson and I have to use ninja. But guess 
what, ninja doesn't accept --timeout-multiplier (nor does its -h output 
suggest something of that kind). I'd understand if one was "user 
friendly" interface of the other, but that doesn't seem to be the case.
I guess I'm annoyed with having to use different tool each time.

Michal

On Fri, Nov 20, 2020 at 08:10:57AM +0100, Michal Privoznik wrote:
> On 11/13/20 8:08 PM, Andrea Bolognani wrote:
> > On Fri, 2020-11-13 at 16:58 +0100, Michal Privoznik wrote:
> > > On 11/8/20 10:24 PM, Roman Bolshakov wrote:
> > > > -    - if test "$(uname)" = "FreeBSD"; then ninja -C build dist; fi
> > > > -    - if test "$(uname)" = "Darwin"; then ninja -C build && ninja -C build install; fi
> > > > +    - ninja -C build dist
> > > 
> > > If we go with --timeout-multiplier= as I'm suggesting in 3/4 then this
> > > can't be ninja, but meson. I'm not sure what the whole point of 'ninja'
> > > is at this point, sorry.
> > 
> > ninja is the low-level tool, so in some cases (notably running the
> > test suite) having meson call ninja instead of invoking the latter
> > directly can enable additional features.
> 
> And I guess that's my problem with it. The --timeout-multiplier= is argument
> of meson, but if I want to run a single threaded build (useful if I mess up
> something in a header file that's included from everywhere) there's no way
> to pass "-j1" to meson and I have to use ninja. But guess what, ninja
> doesn't accept --timeout-multiplier (nor does its -h output suggest
> something of that kind). I'd understand if one was "user friendly" interface
> of the other, but that doesn't seem to be the case.
> I guess I'm annoyed with having to use different tool each time.
> 
> Michal
> 

Hi Michal,

My intention with the patch was to remove the difference (skipping tests
on macOS) that is no longer needed. FreeBSD had the most concise
invocation, so I kept it.

It's not trouble for me to update it to "meson dist" [1] if you think
it's more flexible (available since 0.52.0).

1. https://mesonbuild.com/Creating-releases.html

Thanks,
Roman

On Mon, 2020-11-23 at 16:46 +0300, Roman Bolshakov wrote:
> My intention with the patch was to remove the difference (skipping tests
> on macOS) that is no longer needed. FreeBSD had the most concise
> invocation, so I kept it.
> 
> It's not trouble for me to update it to "meson dist" [1] if you think
> it's more flexible (available since 0.52.0).

The GitLab CI configuration runs 'ninja dist' (as well as 'ninja
test'), so let's stick with that one. We might want to change those
across the board, but that would be a completely separate patch :)

-- 
Andrea Bolognani / Red Hat / Virtualization

On 11/23/20 2:46 PM, Roman Bolshakov wrote:
> On Fri, Nov 20, 2020 at 08:10:57AM +0100, Michal Privoznik wrote:
>> On 11/13/20 8:08 PM, Andrea Bolognani wrote:
>>> On Fri, 2020-11-13 at 16:58 +0100, Michal Privoznik wrote:
>>>> On 11/8/20 10:24 PM, Roman Bolshakov wrote:
>>>>> -    - if test "$(uname)" = "FreeBSD"; then ninja -C build dist; fi
>>>>> -    - if test "$(uname)" = "Darwin"; then ninja -C build && ninja -C build install; fi
>>>>> +    - ninja -C build dist
>>>>
>>>> If we go with --timeout-multiplier= as I'm suggesting in 3/4 then this
>>>> can't be ninja, but meson. I'm not sure what the whole point of 'ninja'
>>>> is at this point, sorry.
>>>
>>> ninja is the low-level tool, so in some cases (notably running the
>>> test suite) having meson call ninja instead of invoking the latter
>>> directly can enable additional features.
>>
>> And I guess that's my problem with it. The --timeout-multiplier= is argument
>> of meson, but if I want to run a single threaded build (useful if I mess up
>> something in a header file that's included from everywhere) there's no way
>> to pass "-j1" to meson and I have to use ninja. But guess what, ninja
>> doesn't accept --timeout-multiplier (nor does its -h output suggest
>> something of that kind). I'd understand if one was "user friendly" interface
>> of the other, but that doesn't seem to be the case.
>> I guess I'm annoyed with having to use different tool each time.
>>
>> Michal
>>
> 
> Hi Michal,
> 
> My intention with the patch was to remove the difference (skipping tests
> on macOS) that is no longer needed. FreeBSD had the most concise
> invocation, so I kept it.
> 
> It's not trouble for me to update it to "meson dist" [1] if you think
> it's more flexible (available since 0.52.0).

My e-mail was more of a rant than suggestion for you to change anything. 
Well, I guess every build system is all sunshine and roses until you 
start using it.

Michal