Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

Loading…

signup flow sometimes leads to 'aw snap' chrome problems #2691

Closed
6a68 opened this Issue · 18 comments

4 participants

@6a68
Owner

First noticed this via selenium tests, interestingly.

Basically, go through the flow of the remove-email automation test:

  1. go to 123done, use the dialog to add an unverified secondary email. don't close the dialog.
  2. pull the link out of the email and paste it into the 123done main window
  3. aw, snap!

Something to do with reusing that window, instead of opening a new tab, causes trouble.

Seen repeatedly in vista/chrome via sauce, for example: https://saucelabs.com/tests/7f0fb60aa7c54f26b35dc30e1731b972

Repro'd manually on mac/chrome stable.

EDITED to add simpler repro steps.

@6a68
Owner

forgot to mention - this happens on dev

@6a68
Owner

NOT occurring on prod

@6a68
Owner

NOT occurring on stage

@jrgm
Collaborator

Five star.

@6a68
Owner

ONLY happens if the dialog is open. If the dialog is closed before loading the link, no error.

@6a68
Owner

not reproducible on mac/ff16

@6a68
Owner

not broken on safari 6 either

@lloyd

alternate STR which are interesting

  1. go to dev.123done.org, use the dialog to add an unverified secondary email. don't close the dialog.
  2. type https://login.dev.anosrep.org into the 123done main window
  3. aw, snap!
@jrgm
Collaborator

So:

1) I have chrome 21 and 22 on osx and can repro with both.

2) Stage does not reproduce we know that. But the same branch on an awsbox does. So what's up with that?

@jrgm
Collaborator

Also on my awsbox, I set options "kpi_backend_sample_rate": 0.0, "enable_development_menu": false which otherwise would alter client behaviour. These are set this way on stage. Any others? (I'd really like to see show_config from stage).

@6a68
Owner

cc @gene1wood, any pointers on config differences?

@jrgm
Collaborator

I had noticed that the 21 and 22 .app's had been updated sometime in the 3 hours between when I restored these from Time Machine and when I noticed they had changed. So I disabled automatic updating and used a clean unupdated chrome 22, and can repro with that. I used sudo lsof -P -n to confirm that there was no version 23 bits loaded anywhere.

So this issue with Chrome may have been around for this flow for quite some time. But for reasons that are unclear, the bug is not reproducible in the stage or production environments which have slight differences in packaging and configuration, and also have a full set of translations.

@6a68
Owner

5-star unassigned, I call shenanigans.

Either this isn't really that important, and we should change our sauce tests to do something different, or it is important, just not 5-star important, or it really is 5-star important, yet it's unassigned and hanging around.

@lloyd, @jrgm, which is it?

@gene1wood
Owner

@6a68 production.json ( https://github.com/mozilla/browserid/blob/1173c2d7c173e9deabe9209b4136006935f1bfaa/config/production.json ) differs from what's in stage in these ways :

  • dev has no public_url
  • dev defines a "user" value under "database"
  • dev has create_schema true under "database"
  • dev has a "kpi_backend_sample_rate" value
  • dev has no "smtp" value
  • dev has a "supported languages" value
  • dev has a express_log_format of default_bid instead of default
  • dev binds the keysigner to 62600 instead of 62700
  • dev has a different verifier url keysigner url and dbwriter url (understandably)
  • dev has a enable_development_menu value
@jrgm
Collaborator

Thanks @gene1wood. The differences for database, smtp, port, express_log_format, {keysigner,dbwriter} url should not matter as this happens on client side. I've set kpi_backend_sample_rate and enable_development_menu to the values they should have in stage (0.0, false) and cannot repro on an awsbox (which is like dev without those two config changes).

I guess what I really need is a run of a command on stage that we can talk about tomorrow.

@6a68 6a68 referenced this issue from a commit
@6a68 6a68 Revert "work around aw snap in remove-email test issue #2691"
  This reverts commit 6a6cf59.

  The about:blank fix causes rampant destruction on FF and Chrome , reverting
  for the moment. We'll find a different fix when we focus on Chrome next week.
a208631
@6a68
Owner

should we reduce the star count, since this only seems to affect dev?

@jrgm
Collaborator

Lowering to 4-star. What @6a68 said.

@6a68
Owner

hasn't recurred in a while, I'm guessing it was a chrome bug. If not, we'll reopen.

@6a68 6a68 closed this
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.