JuliaIO / BufferedStreams.jl

Fast composable IO streams
MIT License
42 stars 20 forks source link

Add Project.toml #60

Closed mcabbott closed 2 years ago

mcabbott commented 4 years ago

This package constrains Compat.jl to version 2, blocking the installation of version 3. I think is assumed from the REQUIRE file, the limit in the registry is here: https://github.com/JuliaRegistries/General/blob/master/B/BufferedStreams/Compat.toml See https://github.com/JuliaRegistries/General/pull/8297 for discussion of other packages with similar issues.

This PR adds the simplest possible Project.toml file, with looser bounds. Perhaps it should also remove the REQUIRE file.

I can't seem to get tests to run locally, but we'll see what travis thinks?

kescobo commented 4 years ago

That's great, thanks! Should probably go ahead and remove REQUIRE as well since it's no longer needed.

mcabbott commented 4 years ago

Done. But lots of test failures, I'm afraid.

DilumAluthge commented 4 years ago

@mcabbott I think it is fine to drop support for Julia 0.7.

So I would recommend that you make the following changes.

First, in Project.toml, please change the line that says julia = "0.7, 1" so that it says julia = "1" instead.

Second, in Project.toml, please add the following lines at the bottom of the file:

[extras]
Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"

[targets]
test = ["Test"]

Third, in .travis.yml, please delete line 6, i.e. the line that looks like this:

  - 0.7
mcabbott commented 4 years ago

Thanks! How did I manage to forget how these files work... and that git add * skips . files... Have also dropped 0.7 as suggested.

DilumAluthge commented 4 years ago

@mcabbott You still need to remove 0.7 from the list of Julia versions being tested in .travis.yml.

codecov[bot] commented 4 years ago

Codecov Report

Merging #60 into master will increase coverage by 28.33%. The diff coverage is 0%.

Impacted file tree graph

@@             Coverage Diff             @@
##           master      #60       +/-   ##
===========================================
+ Coverage   57.23%   85.57%   +28.33%     
===========================================
  Files           5        5               
  Lines         304      305        +1     
===========================================
+ Hits          174      261       +87     
+ Misses        130       44       -86
Impacted Files Coverage Δ
src/BufferedStreams.jl 77.77% <0%> (+15.27%) :arrow_up:
src/bufferedoutputstream.jl 72.15% <0%> (+24.05%) :arrow_up:
src/emptystream.jl 70.58% <0%> (+27.45%) :arrow_up:
src/bufferedinputstream.jl 96.93% <0%> (+31.9%) :arrow_up:

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update b85ff82...9223313. Read the comment docs.

mcabbott commented 4 years ago

I think I can delete the Compat dependence entirely, in fact.

There are problems with tests timing out on 1.3, although locally OK...

DilumAluthge commented 4 years ago

The tests pass on 1.3 on Linux but timeout on macOS. So it is almost certainly a timeout and not a true fail.

Can you add the following line to .travis.yml? It should fix the timeout issue.

script: travis_wait 30 julia --code-coverage -e 'using Pkg; Pkg.build(); Pkg.test(coverage=true)'
mcabbott commented 4 years ago

Yes it passes locally on a mac, the travis machines must just be slower than I guessed.

DilumAluthge commented 4 years ago

The Travis macOS machines are notoriously slower than their Linux machines.

DilumAluthge commented 4 years ago

Travis times out after 50 minutes, so there isn't really a point to putting travis_wait 60.

mcabbott commented 4 years ago

OK, let's hope 50 is enough.

Is it expected that Julia 1.3 is 5 times slower here than 1.0, btw?

DilumAluthge commented 4 years ago

It is certainly concerning.

DilumAluthge commented 4 years ago

Can you test on nightly as well so we can see how fast that is?

mcabbott commented 4 years ago

On travis it's similar to 1.3, the mac one must have been just under the 30min timeout.

mcabbott commented 4 years ago

Bump?

Does anything else need doing here? travis_wait 30 isn't enough to pass reliably, even if 60 isn't taken literally it seems to allow more time. And the slowdown on 1.3 seems like an orthogonal problem.

bjarthur commented 4 years ago

would be nice to have this merged and tagged as it's not having it is causing install conflicts.

mcabbott commented 4 years ago

The registry PR https://github.com/JuliaRegistries/General/pull/11257 ought to have solved the problem of install conflicts.