Conversation
Bump up to 0.36.12? |
Nice catch! No idea that existed. |
Archlinux, builds fine and atom operates as usual. |
The failing spec seems to be flaky on master so it can probably be ignored. |
Friends don't let friends let flakey tests stay flakey 😆 |
Atom is still broken on master without this PR. Same as in another PR, can we get this merged asap? |
what are the failing checks referring to? |
Months of waiting simply to update electron. |
@50Wliu you mentioned that the failing spec seems to be flaky on master as well, does this need to be rebuilt, or can we just approve this PR back to master? I'm wondering if there is anything that we can do to speed this process up |
What can be done to help speed the process up is if people can give us an idea of what testing they have done to ensure that this doesn't create any regressions. We all want to update to the latest version but only the Atom team is on the hook if a regression is introduced. Since Electron can affect literally anything from Atom's UI to how it interacts with the underlying OS to what JavaScript features are available, any Electron upgrade is potentially a breaking change. So, if people want to help, here's what you can do:
If you run into any problems, get steps to reproduce the problem and verify that the problem does not exist on the latest If you do not run into any problems, add a link to a detailed report here (you can store the report in a Gist) of the things you did including:
|
I went as far as updating, locally, electron to 0.37.5 (however all details/findings apply to current PR 0.36.12) and in my 2 days of testing I noticed 0 regressions on ArchLinux, Gnome-Shell running under X. Main workflow consists of developing websites with AngularJS with a couple of extra community packages. Building latest from master, running 1.9.0 DEV atom version and the build is successful. Only changes made were edits to package.json to upgrade electron to 37.5 and changes listed in this PR. script/test shows no problems apart from
But that's 99% because of the fact that I am using Arch and not a debian based system hence the mkdeb issue. My Confg.cson
Packages installed:
I hope this helps, I do understand that the electron in question is 0.36.12 but I do believe there is no active regressions (at least in general sense of use) to update electron to 0.37.5. |
I've been running this patched branch for a few days now. Running smoothly so far. I'll attach config and logs after some more testing. |
*Mac and Windows
@50Wliu not sure if these are related to Electron at all but on stable, packages and themes are fine. Just giving you a heads up just in case 😄 |
@Nikki1993 Please stick to the branch as implemented. We're not looking at updating to Electron v0.37.x yet. Let's constrain testing reports on this PR to ones that actually test this PR. |
@lee-dohm understand. I was testing this PR for the past week and no issues arise, using 0.36.12 electron. Everything what I wrote for 0.37.5 electron applies to 0.36.12 no regressions that I can find that would disturb the workflow. |
As someone who used to work in QA, the bolded text doesn't fill me with confidence that no regressions were actually found 😉 |
@lee-dohm again, there is only so much that can be found. I tested atom from the point I am using it and everything is flawless. I try to go above and beyond, but priority atm is to get atom in a working condition for Arch users. I always leave a seed of doubt, hence the
Naturally, I tend to test software extensively when the time allows but presently atom is broken on Arch based systems and causes headaches to update. I feel like if no problems are found that would break things even further, it's safe to merge and deploy. This PR is a successor to different PR which has been around for a few weeks already. P.S. I am interested in QA testing and I support that a great deal of tests should be executed before merging framework changes, however when things are broken I believe they can be sped up a little bit :P |
I appreciate where you're coming from and, in general, agree with your point of view. What we're looking for is "regressions". My apologies for assuming that everyone is familiar with the term. A regression is defined as something that used to work that no longer does. So yes, we're on the same page. What confused me about your phrasing is that it sounded like you had found regressions (again, things that once worked that no longer do) but had discounted them as not disruptive to some workflow. Since exhaustive testing is generally not possible in any non-trivial system (and definitely not possible in one of Atom's size), any regression is cause for concern. But from your latest message, it sounds like you didn't find any regressions ... just things that haven't ever worked on Arch? |
When you say "latest master", are you referring to Atom's latest master? Or the atom-editor Arch Linux package's latest master? |
@lee-dohm latest Atom's master from git. I build atom-editor myself, not through Arch AUR system. |
However, julia-client is NOT crashing on Atom 1.6.2 Windows 10 64bit. |
Ok, since it is from the latest master, please open a separate issue for it. |
@lee-dohm |
The fact that atom-editor is (only partially) patching in this PR in is extremely concerning. @amytruong I've actually been experiencing that for a while now (at least for two packages) on master. |
@50Wliu Sorry for the confusion, I've checked the patch that the AUR package is actually using and it does not look like the whole PR (however, I have really no idea how it should look). OTOH @Nikki1993 reports the independent build has the same issue. |
Upon looking closer at @Nikki1993's error, that one doesn't seem to be due to module require problems and seems to me like something is wrong with |
This test should have failed prior to the path updates, but it didn't, maybe because this expect doesn't have an assertion (eg @amytruong I'm seeing the |
The aforementioned spec has been fixed and I found a bug as a result 😄. |
I was trying to build AUR package using the patch from this branch (not including the changes from the last hour) instead of the |
@alyst The 1.7.0-4 package now contains all of this PR which went up about 15 minutes ago. The nodegit update is a i686 specfic issue patch unrelated to this PR, you can remove that change and see if it helps your issues. |
@sjug Thanks, 1.7.0-4 so far is working fine for me. |
All works here. Everything in
It is great to have a working Atom again! |
There's no need to run |
@50Wliu what is all the output then? |
As I briefly described in #11516, that's just Atom building itself and then making sure it can publish a release. Currently only OS X runs the specs when |
Thanks @50Wliu and @drewmnoel! 🎉 |
Continuation of #10983.
Closes #10983.
Fixes #11198.
/cc @drewmnoel