Hi Pablo,
Thanks for your quick reply.
On 16 March 2018 at 14:38,
[hidden email] <
[hidden email]> wrote:
> Hi,
> sorry my mistake when updating the scripts. As the bootstrap.sh is not
> executed during the Jenkins process it has not been validated (locally I'm
> not using that script, because is trying to download each time the VM and
> the bootstraping process image)
Right. Can we make one of the goals of the refactoring to avoid any
code duplication?
e.g. at the moment (ignoring changes proposed below), Jenkins should
be calling bootstrap.sh.
> The correct line is the one from the JenkinsFile, it should be something
> like:
>
> ./pharo ./Pharo.image bootstrap/scripts/bootstrap.st --ARCH=${architecture}
> --BUILD_NUMBER=${env.BUILD_ID} --quit
>
> I will submit a PR to fix it, as soon as the CI infrastructure is less on
> fire.
> Until the fix, if you are building the image locally you can change the
> error line with:
>
> ./pharo ./Pharo.image bootstrap/scripts/bootstrap.st --quit
>
> And it will work.
Shall do.
> The idea is to migrate more parts of the process to pure Pharo code,
+1
> everything is implemented in a project PharoBootstrapProcess, which is
> located in bootstrap/src in the same Pharo repository.
> The best way to check what is there is to do the same as the process.
> 1) Take an image with VMMaker, Espell and Ficus (we have them already made
> in
>
https://github.com/guillep/PharoBootstrap/releases/download/v1.4/bootstrapImage.zip).
> 2) Execute bootstrap/scripts/prepare_image.st
>
> That is the same image used to bootstrap Pharo.
> I will do another pass on this first part of the process and generate some
> documentation about the changes and how to modify the whole process.
> We decided to migrate the process by parts to be able to review it easily
> than waiting to have a whole migration of the process.
>
> The next step is to do the same with the build.sh file, as you can see it,
> is a little complex and have lots of steps.
> The idea is again convert them into Pharo code.
Can we also re-order the bootstrap and break it into several distinct phases.
1. Clean up (this ensures a clean environment for Jenkins)
2. Download all pre-requisites.
The main change here is moving the call to download_vm.sh out of build.sh.
This would allow a particular VM to be installed by hand for step
4, see below.
3. Initial preparation, i.e. all the execution prior to build.sh, e.g.
prepare_image.st.
Currently this is done with the Pharo 6 VM and image, which makes sense as it
provides stability.
4. Main image bootstrap. Which is done with the Pharo 7 VM (currently
what is in build.sh).
That will allow the first three steps to be run once and then skipped
when doing personal development / testing.
I want to be able to manually load the VM to be used in step 4 as part
of some stability testing so we can promote a new Pharo 7 VM to
stable.
Are you open to me submitting some PRs that do this restructure to the
phases listed above and removing duplicate code from Jenkinsfile?
> Again sorry if the partial change made some problems in your local execution
> of the bootstrap.
No problem.
Thanks,
Alistair
> Cheers.
>
>
> On Fri, Mar 16, 2018 at 2:02 PM, Alistair Grant <
[hidden email]>
> wrote:
>>
>> Hi Pablo & Everyone,
>>
>> The bootstrap process seems to be in an inconsistent state at the
>> moment as the scripts in:
>>
>> - Jenkinsfile and
>> - bootstrap/scripts/bootstrap.sh
>>
>> don't match.
>>
>> bootstrap/scripts/bootstrap.sh has an error in it:
>>
>> ./pharo Pharo.image eval "PBBootstrap fromCommandLine bootstrap" --quit
>>
>> fails because the "--quit" option is valid for the st handler
>> (executing a smalltalk file), but isn't recognised by the eval
>> handler.
>>
>> Can someone, Pablo?, confirm what the status is, what it should be and
>> what the next steps are? :-)
>>
>> Thanks,
>> Alistair
>>
>
>
>
> --
> Pablo Tesone.
>
[hidden email]
Locked
