Description

At the moment, almost all of our data from Proton users is put together entirely by hand. This is fine if you know the whole process in advance, but if you don't, the process is extremely difficult to understand:

• Add PROTON_LOG=1 %command% as a launch option
• Run game
• Check appid (somehow), search for steam-{appid} in $HOME folder
• Open https://github.com/ValveSoftware/Proton/issues
• Search in the filter texbox for {appid}
• Append report to existing report, or generate a new one
  • New reports get templates, appends are just expected to post... something?
• Check all the tickboxes in the template
• Write in various entries by hand (including GPU/kernel info)
• Go to Steam client, Help > System Information (which has GPU/kernel info)
• Ctrl+A-Ctrl+C
• Oh yeah did I mention, make a GitHub account
• Make a new file at gist.github.com, Ctrl+V
• Add Gist link to template where specified
• Upload steam-{appid} to template where specified
• Describe any notable issues
• Post

16 nonlinear steps and kisak still has to ask people for certain system info, ask for the Proton log, merge issues into other issues, rename/tag issues, etc.

Most of this could be automated by the Steam client, and in fact, a lot of this is linearized/bulletproofed by ProtonDB:

https://www.protondb.com/

Noteworthy is that the Proton issue tracker has ~1500 entries, while ProtonDB has more than 6000 entries (including native, admittedly), and it's not even an official/endorsed database.

For just report generation, I would propose additions to 2 menus in the Steam client:

Steam Play Settings
	Advanced
		+ [ ] Enable logging for all titles -> export PROTON_LOG=1
	+ Issue Tracker -> https://github.com/ValveSoftware/Proton/issues

Game Properties
	+ Steam Play
		Per-Game Setting (Requires restart of game)
			Compatibility tool
			[ ] Enable logging -> export PROTON_LOG=1
		Issue Tracker -> https://github.com/ValveSoftware/Proton/issues?q=${APPID}
		Generate Whitelist Report
		{
			{NO LOG -> "Please run the game with logging enabled first" -> BAIL}
			[ ] "Pressing the Play button the Steam client is sufficient"
			[ ] "Runtime config options are necessary to run the game"
			[ ] "No workarounds other than "PROTON_* env variables are necessary"
			[ ] "I haven't experienced any issues"
			[ ] "There are no issues left open for this game"
			[ ] "Although I consider the gaming experience equal to Windows there are remaining issues:" -> Textbox
			Generate
			{
				"Report generated and copied to clipboard."
				"Be sure to upload $HOME/steam-${APPID} into your report!"
			}
		}
		Generate Compatibility Report
		{
			{NO LOG -> "Please run the game with logging enabled first" -> BAIL}
			[ ] "I have checked whether there are updates for my system available"
			Generate
			{
				"Report generated and copied to clipboard."
				"Be sure to upload $HOME/steam-${APPID} into your report!"
			}
		}

This all assumes that we add the system info as plain text to the report (maybe we would use a collapsed box for that?), still expect manual uploads of logs, and only half-automate appid searching in the tracker. This at least solves the first half of the problem (generating the data in the first place), but the uploading part is still really crappy. It leaves two big questions:

1. Can we automate picking the right game in the issue tracker?
2. Can we automate uploading the log files when generating a post?

This part is out of my range, but it'd be nice if we had something that reduce report generation to a handful of clicks rather than what we have now...

Justification

We've gathered a lot of data, but our capacity for gathering more data competently is really really low. It puts a lot of work onto both users and developers, which is a bad customer experience for the former and a waste of valuable resources for the latter.

Additionally, if we generate the data in the client, it makes the data easier for a program to understand, meaning we can automate certain aspects like system requirement checks, driver version checks, and tagging (search for XAudio2 references, for example).

Risks

Because we still use GitHub, this still means that reporting requires an account aside from the Steam account. I'm sure GitHub doesn't mind having an inflated user count, but a general audience may find this questionable. ProtonDB uses Steam logins for accounts, which is nice!