Support ANSI color codes
It would be great if the log output would support ANSI encoding for colors. Currently we have to disable it but then we loose a lot of information (the one glance green=good - red=bad).
I'm used to using Travis CI and they manage to encode it to HTML. They support only the colors and style encoding but that is fine, cursor control get to complicated (although there is talk to support it too).
Comments are currently closed for this discussion. You can start a new one.
Keyboard shortcuts
Generic
| ? | Show this help |
|---|---|
| ESC | Blurs the current field |
Comment Form
| r | Focus the comment reply box |
|---|---|
| ^ + ↩ | Submit the comment |
You can use Command ⌘ instead of Control ^ on Mac
Support Staff 1 Posted by Feodor Fitsner on 24 Apr, 2014 10:09 PM
Is it for Node.js output?
2 Posted by Bart van der Sc... on 24 Apr, 2014 10:15 PM
Yes, it is for a node.js build.
Colors are produced by various means, I use Grunt and it uses these https://github.com/marak/colors.js/
To compare:
Appveyor: https://ci.appveyor.com/project/BartvanderSchoor/tsd/build/3
Same on Travis: https://travis-ci.org/DefinitelyTyped/tsd/builds/23711876
Support Staff 3 Posted by Feodor Fitsner on 24 Apr, 2014 10:50 PM
It's Windows and the way it grabs process output. We've been investigating the problem of capturing color information from console but it seems it's too deep inside Windows and we decided to abandon that idea.
Another thing are ESC color sequences - I wouldn't mind to give it a chance as it seems they can be processed on client side in JS.
4 Posted by Bart van der Sc... on 24 Apr, 2014 11:05 PM
Afaik these ANSI codes are simple string codes right in the output stream: you see them in the logs. So should be possible to capture these as-is via stdout right?
Would it matter if I use a different shell? I have
ps: grunt test(PowerShell) but for no particular reason.Support Staff 5 Posted by Feodor Fitsner on 24 Apr, 2014 11:09 PM
Yeah, I think ANSI colors can be easily processed on client side.
Switching between shells won't make any difference - it's still running Node.js in a separate process and capturing output.
6 Posted by Bart van der Sc... on 24 Apr, 2014 11:23 PM
Cool, will be nice.
If you start doing client side stuff on the logs then you can some other interesting things: like making toggle-groups from boring parts of the builds and hiding them after they are done. (like npm-install is so boring to scroll through)
Support Staff 7 Posted by Feodor Fitsner on 24 Apr, 2014 11:25 PM
Yep, grouping would be nice too.
Support Staff 8 Posted by Feodor Fitsner on 24 Apr, 2014 11:50 PM
Well, well, it seems I've got some results ;)
9 Posted by Bart van der Sc... on 25 Apr, 2014 12:02 AM
Hah, nice.
:shipit:
Support Staff 10 Posted by Feodor Fitsner on 25 Apr, 2014 01:08 AM
Deployed!
- Feodor
11 Posted by Bart van der Sc... on 25 Apr, 2014 03:24 AM
Very cool, nice colors too.
I did spot a weird edge-case:
My build runs a node.js child process that streams back to the main process. (using
grunt-shell). For some weird reason it doesn't get filtered properly:Compare this: https://ci.appveyor.com/project/BartvanderSchoor/tsd#L1214
To: https://travis-ci.org/DefinitelyTyped/tsd#L1108
(also note the log-line-deeplink to appveyor doesn't offset the scroll-position for the height of the page header :)
Support Staff 12 Posted by Feodor Fitsner on 25 Apr, 2014 03:27 AM
Yeah, that fixed header is a pain :)
So, what's the difference between Travis and AV logs? :-/
13 Posted by Bart van der Sc... on 25 Apr, 2014 03:43 AM
It didn't show? Weird.
Check the attached screenshots: notice AppVeyor has a lot more cyan then Travis.
Could be because of chunk boundaries in the stdio streams?
Support Staff 14 Posted by Feodor Fitsner on 25 Apr, 2014 04:34 AM
Oh, I see. Thanks for the catch!
It's been incorrect handling of "reset"
[0msequence. It was fixed and updated deployed.15 Posted by Bart van der Sc... on 25 Apr, 2014 09:30 AM
I see it, very nice! (and good bright colors too).
16 Posted by tony on 25 Apr, 2014 10:46 PM
Nice, glad colors are now implemented. The old [1m*[0m [31m was looking a bit messy.
17 Posted by Kirill Müller on 14 Aug, 2014 10:13 AM
Thanks. Is it possible to colorize my own output with the Write-Output cmdlet in PowerShell?
Support Staff 18 Posted by Feodor Fitsner on 14 Aug, 2014 03:20 PM
You can have colorized output with
Write-Host19 Posted by Kirill Müller on 14 Aug, 2014 04:37 PM
Works like a charm! For reference:
WriteHost "message" -ForegroundColor Green20 Posted by nathan.stocks on 22 Jun, 2015 07:53 PM
This isn't working for me at all. What am I missing?
My project uses colorama for colors on Windows. It works great on real Windows, but it doesn't seem to work at all on AppVeyor.
Here's a build that should be very colorful:
https://ci.appveyor.com/project/CleanCut/green/build/job/5k9esr3b04fwyax5
21 Posted by nathan.stocks on 22 Jun, 2015 10:34 PM
I should probably clarify that colorama is a python module that handles portable terminal color codes (posix and windows).
Support Staff 22 Posted by Feodor Fitsner on 23 Jun, 2015 02:10 AM
AppVeyor can only display whatever goes to STDOUT/STDERR of the process. I guess those libraries work with console API.aspx) directly.
23 Posted by nathan.stocks on 23 Jun, 2015 02:27 AM
Are you saying that instead of (colorama) stripping the ansi escape
sequences and making the win32 calls directly, it should actually output
the ansi escape sequences that typically don't work on windows -- but that
appveyor will convert them?
Here's the python library I'm using, for reference:
https://pypi.python.org/pypi/colorama
Support Staff 24 Posted by Feodor Fitsner on 23 Jun, 2015 02:44 AM
I'm not an expert in Python but looking into this source file proves that the library uses Console API. Unfortunately, this information cannot be (easily) intercepted while running the process. Obviously, as ANSI sequences go to STDOUT they can be interpreted by AppVeyor.
25 Posted by nathan.stocks on 23 Jun, 2015 03:56 AM
Thanks! With that information I was able to simply *not* turn on Windows
ansi coloring iff we are running on Windows on Appveyor...to turn on
colors. :-)
https://ci.appveyor.com/project/CleanCut/green/build/job/5a01wnkd58gt2ms2
26 Posted by Nicolas on 20 Aug, 2015 09:55 AM
Hello, I'm trying to use appveyor to test the Symfony PHP framework on Windows.
I'm almost done, you did a wonderful job here!
I wanted to notice you of a bug in the ansi color support:
Only single codes are supported ( \e[1m ). Multiple codes sequences are not handled ( \e[1;2;3m ). See e.g. https://ci.appveyor.com/project/fabpot/symfony/build/1.0.72
I worked around this current limitation by splitting multiple codes sequences to single codes, but then sometimes, colors are not correctly reseted, see e.g.:
https://ci.appveyor.com/project/fabpot/symfony/build/1.0.74#L1256
27 Posted by Nicolas on 20 Aug, 2015 10:12 AM
Also not that sometimes you split a log line in the middle of an ansi code (see raw log), leading the display bugs
e.g. https://ci.appveyor.com/project/fabpot/symfony/build/1.0.74#L1022
Support Staff 28 Posted by Feodor Fitsner on 20 Aug, 2015 11:06 AM
@Nicolas What's the meaning of those multi-code sequences? Could you drop a link to more info about it?
29 Posted by Nicolas on 20 Aug, 2015 11:43 AM
For almost all codes, this is just a way to concat several codes in one. Which means \e[1;2;3m is the same as \e[1m\e[2m\e[3m
There are two exceptions: modes 38 and 48, which consume some numbers as parameters (but I suggest you don't implement them, at least for now, if that's too much work):
https://en.wikipedia.org/wiki/ANSI_escape_code#graphics
30 Posted by Nicolas on 29 Sep, 2015 08:14 AM
https://github.com/appveyor/ci/issues/373