#github-actions

2021-08-31

Kevin Rue-Albrecht (07:29:19): > @Kevin Rue-Albrecht has joined the channel

Kevin Rue-Albrecht (07:29:19): > set the channel description: Questions and advice for running GitHub actions for Bioconductor packages

Kevin Rue-Albrecht (07:29:47): > Anyone else recently ran into this one? > > Error in loadNamespace(i, c(lib.loc, .libPaths()), versionCheck = vI[[i]]) : > there is no package called 'SummarizedExperiment' > Calls: <Anonymous> ... loadNamespace -> withRestarts -> withOneRestart -> doWithOneRestart > ERROR: lazy loading failed for package 'rtracklayer' > Execution halted > * removing '/__w/_temp/Library/rtracklayer' > ERROR: dependency 'MatrixGenerics' is not available for package 'SummarizedExperiment' > * removing '/__w/_temp/Library/SummarizedExperiment' > ERROR: dependency 'rtracklayer' is not available for package 'GenomicFeatures' > * removing '/__w/_temp/Library/GenomicFeatures' > ERROR: dependencies 'GenomicFeatures', 'rtracklayer' are not available for package 'ensembldb' > * removing '/__w/_temp/Library/ensembldb' > > ->https://github.com/kevinrue/TVTB/runs/3471391489?check_suite_focus=true#step:17:155

Kevin Rue-Albrecht (07:31:04): > I remember running into errors like > > ERROR: dependency 'MatrixGenerics' is not available for package 'SummarizedExperiment' > > a while back, I just can’t remember what the fix was

Kevin Rue-Albrecht (07:31:23): > it seems that the earliest error in that particular workflow is > > Warning: dependency 'matrixStats (>= 0.60.1)' is not available >

Kevin Rue-Albrecht (07:32:01): > Though it looks like that is a version recently released on CRAN (2021-08-23):https://cran.rstudio.com/web/packages/matrixStats/index.html - Attachment (cran.rstudio.com): matrixStats: Functions that Apply to Rows and Columns of Matrices (and to Vectors) > High-performing functions operating on rows and columns of matrices, e.g. col / rowMedians(), col / rowRanks(), and col / rowSds(). Functions optimized per data type and for subsetted calculations such that both memory usage and processing time is minimized. There are also optimized vector-based methods, e.g. binMeans(), madDiff() and weightedMedian().

Charlotte Soneson (07:39:35): > @Charlotte Soneson has joined the channel

Charlotte Soneson (07:41:14): > It may be that it’s trying to pick upmatrixStatsfrom the RStudio package manager, where it seems that only 0.60.0 is available at the moment (we’re running into exactly this issue in a course at this very moment - specifying thatmatrixStatsshould be installed fromrepos = "https://cloud.r-project.org"worked there).

Luke Zappia (07:49:11): > @Luke Zappia has joined the channel

Charlotte Soneson (08:01:40) (in thread): > The snapshot is from Aug 9, before the newmatrixStatspackage was released:https://github.com/kevinrue/TVTB/runs/3471391489?check_suite_focus=true#step:17:26

Alan O’C (08:32:44): > @Alan O’C has joined the channel

Marcel Ramos Pérez (09:16:46): > @Marcel Ramos Pérez has joined the channel

Spencer Nystrom (09:47:03): > @Spencer Nystrom has joined the channel

Pablo Rodriguez (10:00:25): > @Pablo Rodriguez has joined the channel

David Zhang (10:12:27): > @David Zhang has joined the channel

Lukas Weber (10:43:37): > @Lukas Weber has joined the channel

Lukas Weber (10:45:34): > I have had this issue in the last week too - thanks@Charlotte Soneson@Kevin Rue-Albrecht

Lukas Weber (10:46:39): > I thought it might also be because the Bioconductor docker build hasn’t been updated in the last 1-2 weeks, which I think my GHA workflow was using:https://github.com/Bioconductor/bioconductor_docker

Lukas Weber (10:47:13): > So this also still has the older version ofmatrixStats

Kevin Rue-Albrecht (10:55:06): > Brilliant. Thanks@Charlotte Soneson! > I was going a bit nuts trying to understand it, and haven’t got used to checking the source repositories. > I can see now that Mac is having a better time picking a repository: > > replacement repositories: > CRAN:[https://cloud.r-project.org](https://cloud.r-project.org) >

Federico Marini (12:27:41): > @Federico Marini has joined the channel

Daniela Cassol (12:46:40): > @Daniela Cassol has joined the channel

Lukas Weber (15:16:39): > Thanks@Charlotte Soneson! I have addedoptions(repos = c(CRAN = "https://cloud.r-project.org"))in my GHA workflow now and seems to be working:tada:

Peter Hickey (19:30:57): > @Peter Hickey has joined the channel

2021-09-01

Laurent Gatto (00:11:17): > @Laurent Gatto has joined the channel

Milan Malfait (05:07:05): > @Milan Malfait has joined the channel

Sean Davis (07:02:34): > @Sean Davis has joined the channel

2021-09-07

Chris Vanderaa (06:27:17): > @Chris Vanderaa has joined the channel

Mike Smith (08:17:33): > @Mike Smith has joined the channel

2021-09-09

Davide Corso (08:36:15): > @Davide Corso has joined the channel

2021-10-08

Tim Triche (11:59:53): > @Tim Triche has joined the channel

2021-11-25

Quang Nguyen (11:18:49): > @Quang Nguyen has joined the channel

2021-11-26

Francesc Català (06:46:09): > @Francesc Català has joined the channel

2022-01-03

Kurt Showmaker (17:04:01): > @Kurt Showmaker has joined the channel

2022-02-25

Mike Smith (11:29:14): > Resurrecting a long dormant channel! I got fed up trying to remember to increment the versions of R in my various workflows after each Bioconductor release, so I tried to develop my own action that would do it for me if I specify I want to use Bioconductor “devel” or “release”. I thought it might be useful to others who are using Github Actions to test packages, so you can find it athttps://github.com/grimbough/bioc-actionsI’d welcome feedback if anyone does try it out.

Alan O’C (11:45:43) (in thread): > Huge thanks, I have sighed at my github actions a few times and thought “I should really just do this”

2022-02-27

Charlotte Soneson (13:14:55) (in thread): > @Mike Smiththis is great, thanks for putting it together!:slightly_smiling_face:May I ask whether you have tried it out also in the Bioc docker containers? I got an error that seems to happen during the tex installation/update, not sure whether it’s something transient so I just wanted to check (seehttps://github.com/csoneson/alevinQC/runs/5351147958?check_suite_focus=true).

2022-02-28

Mike Smith (04:10:27) (in thread): > No, I’ve not tried using the docker containers. Personally, while I think the Docker images are great for getting things done, when it comes to checking code I actually like to have a record of the required system packages etc so I can help users with installation problems. Hence no docker testing yet. > > Looking at that log, it seems my use of Powershell to run thetlmgrfails as Powershell isn’t available in the container. That step is probably only required on Windows, and definitely not in the container, so I’ll try to make it conditional.

2022-03-01

Mike Smith (03:52:09) (in thread): > ~~~I think I’ve got it working with Docker now. I’ve got a small example workflow now at ~https://github.com/grimbough/biocActionsExamples/runs/5372075755?check_suite_focus=true~~ which seems to be passing. I don’t think it’s perfect, but it’s working for now. ~~~~Thanks for the feedback. > > Scratch that. I made an assumption that$R_USER_LIBSwould be defined, but it isn’t. I don’t understand why I get a green tick even thoughR CMD checkfails, but will investigate.

Charlotte Soneson (04:20:55) (in thread): > Thanks! Just tried it in my repo as well and I get the same warning in the check (but a green checkmark)

Mike Smith (11:50:04) (in thread): > Well I still don’t understand why it gave a green tick, I guessR CMD checkstill gives an exit code0:shrug:Either way, I think I’ve sorted the library path when working inside a container, so feel free to give it another try.

Charlotte Soneson (12:25:09) (in thread): > Confirmed, no more warning:tada:Thanks!

2022-03-15

Jeanette Johnson (12:03:52): > @Jeanette Johnson has joined the channel

2022-03-18

Andres Wokaty (10:00:28): > @Andres Wokaty has joined the channel

2022-03-25

Charlotte Soneson (04:17:40): > Looks like we’re at this time of the year again…https://community-bioc.slack.com/archives/CLUJWDQF4/p1619684796117500.@Mike Smith- I recall you put something together with aprereleaselast year; do you think that’s something that could be baked into your bioc-actions? - Attachment: Attachment > Question for those that use GitHub actions. We’ve entered the weird limbo period prior to an R release, where R-devel is now R-4.2, but R-release is still 4.0. This breaks all my GH action workflows with Bioconductor version '3.13' requires R version '4.1'; R version is too new; errors. What’re peoples strategies for handling this?

Mike Smith (05:45:46) (in thread): > It’s tricky. Last time I made a copy of ther-lib/setup-raction but that’s not something I really want to maintain. My actions sit on top of that and pass the R version argument, but I don’t think they provide an option to install the pre-release versions. From what I remember the names of the install binaries change really frequently at this time (and might be inconsistent across platforms?) so it’s a nightmare to stay ontop of without something automated to find the names. > > I’ll have a think about what might work.

2022-03-28

Sarvesh Nikumbh (10:42:54): > @Sarvesh Nikumbh has joined the channel

2022-03-29

Dania Machlab (09:05:41): > @Dania Machlab has joined the channel

2022-04-22

Charlotte Soneson (08:44:44) (in thread): > I just noticed thathttps://github.com/r-lib/actions/tree/v2-branch/setup-r(v2) now has the option to specify thenextR version (which should give you the alpha/prerelease etc during the release process, and the patched version otherwise). It will still take some fiddling with the logic to figure out which version we need, but at least now it seems possible to get it.

Mike Smith (08:47:46) (in thread): > Awesome. I’m on leave at the moment, but I’ll take a look when I’m back. Could you open an issue on my GitHub repo so I remember to do this when I’m back?

Charlotte Soneson (08:48:00) (in thread): > Will do!

2022-04-25

Laurent Gatto (10:49:13): > Hello - I have very recently been hit by the following issue on two different repos, claiming that the repo is owned by someone else. The error happens in the pkgdown deployment on the gh-pages branch: > > Run git config --local user.email "action@github.com" > > fatal: --local can only be used inside a git repository > fatal: --local can only be used inside a git repository > > Error: <system_command_status_error in callr::run("git", c(...), echo_cmd = echo_cmd, echo = echo, error_on_status ...: > System command 'git' failed, exit status: 128, stderr: > > E> fatal: unsafe repository ('/__w/PSMatch/PSMatch' is owned by someone else) > E> To add an exception for this directory, call: > E> > E> git config --global --add safe.directory /__w/PSMatch/PSMatch> > > Stack trace: > 1. pkgdown::deploy_to_branch(new_process = FALSE) > 2. pkgdown:::git_current_branch() > 3. pkgdown:::git("rev-parse", "--abbrev-ref", "HEAD", echo = FALSE, ... > 4. callr::run("git", c(...), echo_cmd = echo_cmd, echo = echo, error_on_status ... > 5. throw(new_process_error(res, call = sys.call(), echo = echo, ... > > x System command 'git' failed, exit status: 128, stderr: > E> fatal: unsafe repository ('/__w/PSMatch/PSMatch' is owned by someone else) > E> To add an exception for this directory, call: > E> > E> git config --global --add safe.directory /__w/PSMatch/PSMatch > Execution halted > Error: Process completed with exit code 1. > > Any idea?

Laurent Gatto (10:49:35) (in thread): > Here’s an examplehttps://github.com/rformassspectrometry/PSMatch/runs/6159341714?check_suite_focus=true

Charlotte Soneson (11:05:35) (in thread): > Just a thought, but I wonder if it would help to update the workflow - the current version ofbiocthisis using different git configuration values (see change here:https://github.com/lcolladotor/biocthis/commit/c06e66d60bd864ef67dffb09f1efb2d5179c62de). Not sure if this is the reason -velociraptoruses the new values and deployed ok last time it was run, but it seems that yours did too until recently so I’m not sure…

Laurent Gatto (11:10:29) (in thread): > Indeed, it used to work:cry:. > > I’m now testingthis one(taken from another package that seems to deploy properly). Will try the latestbiocthisversion next. Thanks!

Marcel Ramos Pérez (12:57:47) (in thread): > Why not usehttps://github.com/JamesIves/github-pages-deploy-action?

Laurent Gatto (13:28:31) (in thread): > Mainly because I have been using gha for a long time and have been keeping the same ones as long as they worked. But I’ll looked into that one, thanks.

Marcel Ramos Pérez (13:30:14) (in thread): > See the template here:https://github.com/seandavi/BuildABiocWorkshop/blob/master/.github/workflows/basic_checks.yaml

2022-05-09

Vince Carey (06:21:48): > @Vince Carey has joined the channel

Vince Carey (06:24:37): > If we are going to be using github actions more systemically, it seems to make sense that BiocCheck examine the actions and warn about aspects of workflows that could be stale or improper. Do we have/want project-endorsed actions with versions and their own testing processes?

Vince Carey (06:38:58): > @Marcel Ramos Pérezthoughts about BiocCheck looking at .github/workflows ?

Vince Carey (06:48:02): > @Mike SmithI know you have defined some actions for use in workflows … would we want to have a versioning discipline for these, linked to a container linked to a Bioconductor version? And some kind of routine testing discipline with reporting to action maintainer?

Lori Shepherd (10:15:24): > @Lori Shepherd has joined the channel

Marcel Ramos Pérez (18:55:28) (in thread): > I don’t think the actions are in the scope ofBiocCheck. GA validation could be part of thebiocthispackage.

Vince Carey (18:56:19) (in thread): > OK

2022-05-10

Mike Smith (06:06:36) (in thread): > Personally I see the actions I’ve developed as developer tools that aid with package development and robustness, rather than providing environments for package users to do something with the package itself. > > Others can chime in if I’m wrong, but I think the most common use case for GH workflows is for the package developer to runR CMD checkon all three platforms and get feedback faster than waiting for the Bioc build system to process everything. Similarly, getting things like code coverage reports are neat, but generally focused on feedback to the package developer (perhaps with the side effect of publicly advertising some level of quality via a badge/report). These are all tasks someone could run on their own hardware and Bioconductor would be none-the-wiser, but GH workflows are a nice way to automate and use hardware for free. I’m not sure it’s the place of BiocCheck or the like to actually care what the workflows are doing. Are they even distributed with a source package? I guess we encourage unit tests, which probably fall into a similar category, but we don’t enforce their use nor check how comprehensive or appropriate any existing tests are andtestthattests are not distributed to the end user by default. > > I’ve deliberately tried to make my actions work outside of the Bioconductor versioning, precisely because it was a pain as a developer to go through all of myR CMD checkworkflows (one per package) and update the versions on R and Bioconductor after each release. Generally I want to setup a workflow that says “this should always use the devel version of Bioconductor” and then let the action make the correct choices about versions for that moment in time, that same as when I submit to the Bioc git repo and assume that the build system will use the appropriate versions for each branch. The action should certainly be versioned, in case I choose to make a design decision that breaks the workflows of others, but I’m not sure it’s directly related to Bioconductor’s versioning.

2022-06-07

Simon Pearce (09:28:04): > @Simon Pearce has joined the channel

2022-07-08

Athena Chen (12:00:51): > @Athena Chen has joined the channel

2022-07-13

Simon Pearce (12:00:01): > Is it possible to have a github action that pushes a container containing the package + all dependencies onto DockerHub?

Kevin Rue-Albrecht (13:14:01) (in thread): > have you checkedbiocthisand more specifically that part of the template?https://github.com/lcolladotor/biocthis/blob/master/inst/templates/check-bioc.yml#L303

Sean Davis (13:36:11) (in thread): > I’ve convertedhttps://github.com/seandavi/BuildABiocWorkshopto use GitHub container registry instead of DockerHub, but the actions in the repo do what you want. It is a template that you can use directly and has instructions for use.

Simon Pearce (14:32:50) (in thread): > I am usingbiocthisyes, which is built on top ofBuildABiocWorkshop from what I can tell. > I didn’t have a Dockerfile apparently, just added that in fromBuildABiocWorkshop. > The run of the action is now building the object, just failed to actually push it to Dockerhub. > > denied: requested access to the resource is denied > Error: exit status 1 > Usage: > exit status 1 > github-actions build-push [flags] > Flags: > -h, --help help for build-push > > I don’t really care where the container goes, just that I can get at it! So GitHub container registry sounds fine too. Let me take a further look atBuildABiocWorkshop.

Simon Pearce (14:41:58) (in thread): > Ok, trying the action block for the GitHub container registry that you have in BuildABiocWorkshop now.

2022-07-14

Simon Pearce (04:24:16) (in thread): > Thank you, that seems to have worked, I now have a container on the github container registry. Just need to figure out how to change the permissions on that to be able to download it…

2022-07-21

Nitesh Turaga (15:53:59): > @Nitesh Turaga has joined the channel

2022-07-28

Mervin Fansler (17:19:54): > @Mervin Fansler has joined the channel

2022-08-01

Garth Kong (12:13:52): > @Garth Kong has joined the channel

2022-09-26

Martin Morgan (18:24:48): > @Martin Morgan has joined the channel

Martin Morgan (18:31:19): > I’ve been working on a ‘recreational’ packagehcaCaseStudiesthat has some relatively expensive file download and computation steps in the vignettes, and I’ve been using BiocFileCache to save both the files and the computations. This means that the vignettes build build quickly on my local machine, where the files and results are cached. > > I’d like to use pkgdown and github workflows to build a github pages site, and would like to take the same approach – caching the expensive download and computation steps for use across workflow runs. > > Is that possible? Can I somehow re-use renv in a straight-forward way, like telling BiocFileCache to save files in a particular directory? Other not-too-complicated solutions?

Spencer Nystrom (18:41:06) (in thread): > Hey Martin. I think the magic bullet here is github’sactions/cache(details here:https://docs.github.com/en/actions/using-workflows/caching-dependencies-to-speed-up-workflows). I haven’t done this for R, but you should be able to configure your renv / BiocFileCache setup into a script to write into your actions cache, then run subsequent builds using the cached env.

Spencer Nystrom (18:42:14) (in thread): > Here is the relevant docs section for using renv for this task:https://docs.github.com/en/actions/using-workflows/caching-dependencies-to-speed-up-workflows

Martin Morgan (18:48:23) (in thread): > Thanks@Spencer Nystromthe second link posted is the same as the first?

Spencer Nystrom (18:50:13) (in thread): > ~No, the first links to a page from github action docs describing a bit of detail about how the cache action works. The second is a link to a subheading in the cache action repo README with an example implementation for using renv in the caching action~.

Spencer Nystrom (18:50:22) (in thread): > Oh weird.

Spencer Nystrom (18:50:36) (in thread): > My bad. Clipboard did something silly. 1 sec.

Spencer Nystrom (18:51:51) (in thread): > https://github.com/actions/cache/blob/main/examples.md#r—renv

Spencer Nystrom (18:51:54) (in thread): > Here we go.

Spencer Nystrom (18:52:06) (in thread): > That’s what I get for replying hastily on the phone!

2022-09-27

Vince Carey (07:30:30) (in thread): > Just noting > > GitHub will remove any cache entries that have not been accessed in over 7 days. There is no limit on the number of caches you can store, but the total size of all caches in a repository is limited to 10 GB. > > fromhttps://docs.github.com/en/actions/using-workflows/caching-dependencies-to-speed-up-workflows

2022-10-11

Nitesh Turaga (16:10:37): > Anyone know how to pass a GITHUB_PAT to a docker build github action?

Nitesh Turaga (16:11:03): > My solution looks like this, but it doesn’t seem to workhttps://github.com/r-lib/actions/issues/199

Andres Wokaty (16:23:06) (in thread): > Looking at that, I wonder if it exposes the PAT in the output? If it’s defined as an environment use${{ env.GITHUB_PAT }}or directly with${{ secrets.GITHUB_PAT }}, at least that’s how I’ve used it. I don’t thinkGPAT=${GITHUB_PAT}would work.

2022-10-12

Nitesh Turaga (08:09:57) (in thread): > Thanks@Andres Wokaty! It seems to be completely “masking” the PAT. > > Within the container when I try to look up GITHUB_PAT, its just an empty string"". I must be doing something wrong.

Nitesh Turaga (09:40:07) (in thread): > Went back to the tried and testedhttps://github.com/marketplace/actions/build-and-push-docker-imagesaction. > > Allows build args to be taken in easily. > > This works well.

2023-01-12

Martin Grigorov (13:06:08): > @Martin Grigorov has joined the channel

2023-01-21

Hien (16:02:25): > @Hien has joined the channel

2023-02-06

Ying Chen (21:37:53): > @Ying Chen has joined the channel

2023-02-14

Jeroen Gilis (04:12:11): > @Jeroen Gilis has joined the channel

2023-03-09

Eunice Miriti (10:50:48): > @Eunice Miriti has joined the channel

Leo Lahti (19:10:20): > @Leo Lahti has joined the channel

Leo Lahti (19:12:20): > I encountered a strange issue where the PR passed our GHA checks perfectly, after that failed when building a gitbook:https://github.com/microbiome/OMA/actions/runs/4378436099/jobs/7663176368Does this look familiar: > > error: RPC failed; curl 55 Failed sending HTTP2 data >

Leo Lahti (19:12:44): > > fatal: the remote end hung up unexpectedly > fatal: the remote end hung up unexpectedly > Everything up-to-dateError: Process completed with exit code 1. >

Leo Lahti (19:13:06): > > Error: Process completed with exit code 1. >

Leo Lahti (19:13:59): > I have not been able to troubleshoot this successfully so far.

2023-03-10

Vince Carey (11:02:22): > Are all your secrets up to date? I ran into something like this when my PAT was invalid.

2023-03-21

Y-h. Taguchi (20:10:23): > @Y-h. Taguchi has joined the channel

2023-03-22

Marc Elosua (04:41:11): > @Marc Elosua has joined the channel

Mike Morgan (05:47:27): > @Mike Morgan has joined the channel

Michael Lynch (13:18:58): > @Michael Lynch has joined the channel

2023-03-29

Mike Smith (08:54:01): > I don’t know about anyone else, but I’ve had a few mentions from@Martin Grigorovwho’s leading a project to build Bioconductor packages on ARM64. However, if he encounters a problem, it’s actually pretty hard for a package maintainer to try and debug it unless they also happe to have access to ARM64 hardware. It feels like a some way of emulating a ARM64 environment might be a useful addition to the suite of Github actions. Does anyone have experience trying this with something like Qemu and the Docker images for alternative architectures?

Martin Grigorov (09:07:32) (in thread): > since I’ve been mentioned … :-)

Martin Grigorov (09:08:07) (in thread): > I have few articles explaining how to do this with QEMU or VM

Martin Grigorov (09:08:49) (in thread): > QEMU & Docker:https://martin-grigorov.medium.com/building-linux-packages-for-different-cpu-architectures-with-docker-and-qemu-d29e4ebc9fa5

Martin Grigorov (09:09:29) (in thread): > free VM from Oracle Cloud:https://martin-grigorov.medium.com/github-actions-arm64-runner-on-oracle-cloud-a77cdf7a325ayou can ignore the Github Actions part

Martin Grigorov (09:10:17) (in thread): > the “problem” with Oracle Cloud (and any other cloud provider) is that they ask for Credit Card details even if you use just their Free Tiers

Martin Grigorov (09:11:26) (in thread): > if you use Github Actions thenhttps://github.com/uraimo/run-on-arch-action/is really handy to run your builds and tests on several CPU architectures. But it uses QEMU so it is a bit slow !

Mike Smith (09:18:39) (in thread): > Awesome, thanks for the rundown. I’ll take a look at them.

Martin Grigorov (09:20:48) (in thread): > One more: CircleCI (a competitor of Github Actions) provides Linux ARM64 (and MacOS ARM64) runners for free too. What I like there is that it is possible to connect via SSH to their VMs and debug a failure

Janani Ravi (10:43:07): > @Janani Ravi has joined the channel

Himel Mallick (16:56:35): > @Himel Mallick has joined the channel

2023-03-30

Henrik Bengtsson (10:35:54): > @Henrik Bengtsson has joined the channel

Henrik Bengtsson (10:38:19) (in thread): > Not CI, but it might be useful to mention the M1 mac-builder servicehttps://mac.r-project.org/macbuilder/submit.html

Yikun Jiang (11:42:35): > @Yikun Jiang has joined the channel

Yikun Jiang (11:47:00) (in thread): > Another tips, if you are using apple M1, you can also test the Linux aarch64 with docker:docker run -ti ubuntu bash

2023-04-03

Mengbo Li (20:45:18): > @Mengbo Li has joined the channel

2023-04-04

Jacques SERIZAY (05:02:34): > @Jacques SERIZAY has joined the channel

2023-04-05

Josué Curto Navarro (20:59:45): > @Josué Curto Navarro has joined the channel

2023-04-12

Nils Hoffmann (11:19:08): > @Nils Hoffmann has joined the channel

2023-04-21

Mike Smith (04:33:02): > Is anyone else getting problems with the compbo of R-4.3, pak (or r-lib/setup-r-dependencies) and Mac OS? For many of my packages it looks like the workflow fails to find some Mac binary, then claims to install from source but really installs a Linux binary and falls over e.g.https://github.com/grimbough/rhdf5/actions/runs/4721626812/jobs/8465141595#step:29:147However I don’t see any issues on the pak or r-lib actions repositories, so I’m wondering if it’s just me.

Mike Smith (07:08:37) (in thread): > Update: pak 0.5.0 was released today and seems to have fixed the issue:tada:

2023-05-25

Jacob Krol (17:13:39): > @Jacob Krol has joined the channel

2023-07-01

Ludwig Geistlinger (13:05:39): > @Ludwig Geistlinger has joined the channel

2023-07-13

Brian Schilder (07:02:15): > @Brian Schilder has joined the channel

2023-07-18

Leonardo Collado Torres (12:48:55): > @Leonardo Collado Torres has joined the channel

2023-08-03

Tyrone Lee (10:53:09): > @Tyrone Lee has joined the channel

Ritika Giri (15:57:51): > @Ritika Giri has joined the channel

2023-08-04

Francesc Català-Moll (13:46:46): > @Francesc Català-Moll has joined the channel

2023-08-07

Jiaji George Chen (11:21:24): > @Jiaji George Chen has joined the channel

2023-12-28

David Rach (09:25:31): > @David Rach has joined the channel

2024-02-06

Mike Smith (06:04:55): > Exciting news for those of use who struggle to get things working on Mac ARM64 architecture. GitHub have made M1 runners available on the free tier for GitHub actions:https://github.blog/changelog/2024-01-30-github-actions-introducing-the-new-m1-macos-runner-available-to-open-source/ - Attachment (The GitHub Blog): GitHub Actions: Introducing the new M1 macOS runner available to open source! > GitHub Actions: Introducing the new M1 macOS runner available to open source!

2024-02-07

Alex Mahmoud (14:25:16): > @Alex Mahmoud has joined the channel

2024-03-05

Vamika Mendiratta (02:08:47): > @Vamika Mendiratta has joined the channel

Semango Nicholas (15:20:42): > @Semango Nicholas has joined the channel

2024-03-06

Benedicta Imelda (03:04:15): > @Benedicta Imelda has joined the channel

2024-03-12

Simon Pearce (10:56:48): > Thepackage submission guidelinessay: > 1. Create your own GitHub repository, containing source code structured as anRpackage. The source code must be in the default branch of your GitHub repository. You cannot specify any alternative branches; the default branch is utilized. The default branch must contain only package code. Any files or directories for other applications (Github Actions, devtool, etc) should be in a different branch. > But I see github workflow files in package repositories, such asseuratorplyranges. > Am I missing something here? How do I properly test a repository?

Marcel Ramos Pérez (11:58:42): > Theyshouldbe in a different branch but in practice nowadays the.githubGHA folder can be included indevelbranch. Seurat is probably not a good example as it is a CRAN package but many Bioconductor packages use GHA for testing. We will look at updating the instructions. Thanks!

Lori Shepherd (12:24:10): > Please make sure if anything like this is included that they are listed in a .Rbuildignore file

Simon Pearce (16:32:06): > Thanks. > Ah, didn’t notice Seurat is a CRAN package:wink:, I just assumed it was Bioconductor due to the content.

2024-03-18

Simon Pearce (10:49:07): > Not entirely a github actions question, but not sure where else to ask. > I have a function that retrieves annotation frombiomaRt. > This sometimes fails (particularly on github it seems), and I need to catch this properly, so not to fail the tests. > Does anyone have an example of how to do this (I’m usingtestthat)?

Marcel Ramos Pérez (12:11:26) (in thread): > Create a helper function that tests the connection tobiomaRtand skip the test if unreachable. Seehttps://code.bioconductor.org/search/search?q=skip_if_not.@Mike Smithmay have specific recommendations.

2024-03-19

Mike Smith (05:42:38) (in thread): > Depending on what exactly you’re testing you can use mocking, maybe via packages like {mockery} or {httptest2}. When I’ve used these, I used {mockery} to ‘mock’ the {biomaRt} function, which will bypass actually calling it and just give you something that looks like the expected output. That’s useful if you’re testing your code that processes the function output, but won’t necessarily tell you if {biomaRt} changes what it’s returning. {httptest2} will intercept the actual call to the website and return some content from disk that you store in the package. You can see this in {biomaRt} itself athttps://github.com/grimbough/biomaRt/tree/devel/tests/testthatThat might be more robust if I were to change the {biomaRt} output format from whatever function you’re using. > > In reality both the BioMart server and {biomaRt} are very stable and I don’t expect the structure of the returned values to change at any point, so you’re probably safe picking either of those options. I choose this approach rather than skipping tests if the connection wasn’t available because even when the connection works it can be very slow to talk to the Ensembl server, so if you have many tests running them all can be really frustrating. The mocked versions are basically instananeous once you’ve figured out how to write them.

2024-03-28

Jonas Schuck (14:21:32): > @Jonas Schuck has joined the channel

2024-04-23

Laurent Gatto (09:52:02): > I came across act to run github actions locally. This looks quite interesting to test them and any changes locally first, and reduce the long change-push-wait dev cycle. Just trying/acting on two random localcheck-bioc.ymlfails (probably for good reasons). Has anyone any experience with this?

2024-04-24

Alan O’C (04:26:55) (in thread): > I have tried it, but found that sometimes it doesn’t work immediately and I’ve not been bothered to figure out why over just resorting back to the change-push-wait cycle

2024-08-19

Rema Gesaka (09:40:54): > @Rema Gesaka has joined the channel

2024-10-02

Eva Hamrud (19:06:42): > @Eva Hamrud has joined the channel

2024-10-12

Vince Carey (21:15:04): > Here’s a macos action failing on a libz reference, passing on linux and windows:https://github.com/vjcitn/ontoProc/actions/runs/11274358609/job/31353335313

2024-10-22

Leonardo Collado Torres (14:12:11) (in thread): > Nice, thanks for sharing Mike!

2024-10-31

Mike Smith (08:41:05): > Is anyone else having issues with ther-lib/actions/setup-r-dependencies@v2action? I keep getting failures in the package installation step usingpakbut I can’t figure out why. I don’t see any reports on the GitHub issues relating to this, hence asking here first. For me it seems to fail with different packages, R version, Bioconductor version, OS etc and also sometimes succeeds fine. There error message always seems to be something like: > > Error: > ! error in pak subprocess > Caused by error: > ! `path` must exist > --- > Backtrace: > 1. pak::lockfile_install(".github/pkg.lock") > 2. pak:::remote(function(...) { … > 3. err$throw(res$error) > > Here’s a run with many examples:https://github.com/grimbough/biocActionsExamples/actions/runs/11610124522/job/32328800099

Mike Smith (10:23:17) (in thread): > Answering my own question, it seems settingNcpus: 1stops this behaviour. I guess there’s some race condition that deletes a directory being used by another installation thread when using more cores.

2024-11-08

Federico Marini (10:57:08): > https://github.com/actions/runner-images/issues/10636

Federico Marini (10:57:19): - File (PNG): image.png

Federico Marini (10:57:29): > might affect our GHA workflows, in some manners

2024-11-09

Tim Triche (00:06:45): > Oh ffs Microsoft

Mike Smith (01:26:07): > Have people been using the pre-installed R version in their actions? I think I’ve always relied on eitherr-lib/setup-rorbioc-actions/setup-biocto install a specific version.

2024-11-12

Alex Mahmoud (16:53:02) (in thread): > AFAIK this is true in all GHA workflows widely used, so the only people affected would be folks having done manual setup

2024-11-21

Leonardo Collado Torres (14:28:16) (in thread): > aye, I use r-lib/setup-r

2024-12-12

Hiru (09:44:30): > @Hiru has joined the channel

2024-12-28

Pascal-Onaho (07:56:17): > @Pascal-Onaho has joined the channel

2025-01-10

Federico Marini (07:04:46) (in thread): > https://bsky.app/profile/gaborcsardi.org/post/3lfeu6tzr5c2l

Federico Marini (07:05:09): - File (PNG): image.png

Alex Mahmoud (10:57:59) (in thread): > We’ve discussed this a bit over the past few weeks since the announcement, and as far as we could tell, all widely used Bioconductor actions are either using Bioconductor containers, or already using r-lib actions to setup R, and we don’t seem to rely on pre-existing R in any of our action templates that we could find. If you find an issue with a Bioconductor GHA workflow, please let us know so we update it asap!!

Federico Marini (10:58:40) (in thread): > I did have an issue on a package whose name was different than the repo itself

Federico Marini (10:59:13) (in thread): > that was because the build command gave me something that was not expected to be, and found no archive and failed.

Federico Marini (10:59:40) (in thread): > i solved it by going with a mixture of the biocthis way and something we had in another package

Federico Marini (11:00:08) (in thread): > Thanks for being on top of this! > Probably not everyone uses the Bioc-styled actions but more vanilla ones..

2025-01-14

neslihan oztas (10:36:58): > @neslihan oztas has joined the channel

2025-02-13

JP Flores (13:54:06): > @JP Flores has joined the channel

2025-03-29

Bridget Imarhia (13:12:03): > @Bridget Imarhia has joined the channel