June 04, 2021
On Friday, 4 June 2021 at 18:25:10 UTC, kinke wrote:
> On Thursday, 3 June 2021 at 13:56:32 UTC, Steven Schveighoffer wrote:
>> There is a page that talks about what credits get you:
>>
>> https://docs.travis-ci.com/user/billing-overview/#usage---credits
>
> FWIW, there are unlimited credits for OSS and AArch64/PowerPC/Z architectures, apparently sponsored by ARM and IBM. After the end of Shippable CI, Travis is, AFAIK, the only free CI service offering AArch64 on Linux. And thus still of relevance for LDC.

What are the resources like?

At the time when I tested, Travis's x86_64 offerings clearly weren't meant for any medium-to-large sized project.  Couldn't get anywhere close to finishing the build without hitting the 50 minute timeout, and increasing parallelism would instead mean getting killed for hitting the hard 4GB memory limit.
June 04, 2021
On Friday, 4 June 2021 at 22:10:29 UTC, Iain Buclaw wrote:
> On Friday, 4 June 2021 at 18:25:10 UTC, kinke wrote:
>> On Thursday, 3 June 2021 at 13:56:32 UTC, Steven Schveighoffer wrote:
>>> There is a page that talks about what credits get you:
>>>
>>> https://docs.travis-ci.com/user/billing-overview/#usage---credits
>>
>> FWIW, there are unlimited credits for OSS and AArch64/PowerPC/Z architectures, apparently sponsored by ARM and IBM. After the end of Shippable CI, Travis is, AFAIK, the only free CI service offering AArch64 on Linux. And thus still of relevance for LDC.
>
> What are the resources like?

Still with .org and `arch: arm64`: `nproc` lists either 4 or 32 cores (!), `free` 4 GB (I very much doubt that's hard though). So sometimes a full build succeeds with 8 jobs without out-of-memory retries, yielding an overall job time of about 22 minutes (e.g., https://travis-ci.org/github/ldc-developers/ldc/builds/773502971). At other times with an agent with 4 cores, it's about 45 minutes.

With .com, I've tested their `arch: arm64-graviton2` a while back, which offered only 2 cores, but a significantly higher per-core performance (comparable to x86_64) IIRC.

So not comparable with Shippable, which consistently provided something like 48 cores (with low per-core performance though) and 32/128 GB of memory.
June 12, 2021

On Thursday, 3 June 2021 at 13:56:32 UTC, Steven Schveighoffer wrote:

>

I'm not sure exactly how it works now, I've moved to github actions.

Has anybody written a summary somewhere? AppVeyor is also acting up today and I have to figure out what I need to change.

My needs are simple: Run "dub test" on as many compiler-platform combinations as possible.

Ok, I also like to record the test coverage that seems to be unusual compared to other projects.

June 14, 2021

On Saturday, 12 June 2021 at 10:16:22 UTC, copacetic wrote:

>

On Thursday, 3 June 2021 at 13:56:32 UTC, Steven Schveighoffer wrote:

>

I'm not sure exactly how it works now, I've moved to github actions.

Has anybody written a summary somewhere? AppVeyor is also acting up today and I have to figure out what I need to change.

My needs are simple: Run "dub test" on as many compiler-platform combinations as possible.

TL;DR Use the setup-dlang GH action: https://github.com/dlang-community/setup-dlang

>

Ok, I also like to record the test coverage that seems to be unusual compared to other projects.

Checkout vibe.d's GH Actions CI pipeline for a more elaborate example: https://github.com/vibe-d/vibe.d/blob/v0.9.4-beta.2/.github/workflows/ci.yml

June 14, 2021
On 6/14/21 11:50 AM, Petar Kirov [ZombineDev] wrote:
> On Saturday, 12 June 2021 at 10:16:22 UTC, copacetic wrote:
>> My needs are simple: Run "dub test" on as many compiler-platform combinations as possible.
> 
> TL;DR Use the setup-dlang GH action: https://github.com/dlang-community/setup-dlang

Yeah, I want to mention how absolutely awesome the setup-dlang action is, it makes things SUPER easy to set up!

Thanks for all who made it happen!

-Steve
1 2
Next ›   Last »