:root is potentially cheaper than #main-window, because IDs aren't actually guaranteed to be unique, so ID selectors are treated just like class selectors under the hood. #main-window occurrences: https://dxr.mozilla.org/mozilla-central/search?q=%23main-window+file%3A.css+-path%3Aobj&redirect=false

Review commit: https://reviewboard.mozilla.org/r/69856/diff/#index_header See other reviews: https://reviewboard.mozilla.org/r/69856/

Comment on attachment 8778617 [details] Bug 1292878 - Use :root instead of #main-window; I've pushed this to the try server to take some screenshots: https://screenshots.mattn.ca/compare/?oldProject=mozilla-central&oldRev=6b65dd49d4f045c0a9753ce60bdb4b7b4aaedcf8&newProject=try&newRev=7dc7db5a144bdff04277dd78f9b58f956c1c8fa6 This seems to cause some problems on Windows 7 and 8 presumably because the specificity of the selectors changes. If you leave out the browser-aero.css changes for now, we can land the rest and I can look into fixing browser-aero.css separately.

I half-anticipated that these changes would cause problems on Windows, but I lack access to a Windows machine so I couldn't really confirm. I'll revert the changes to browser-aero.css.

Comment on attachment 8778617 [details] Bug 1292878 - Use :root instead of #main-window; Review request updated; see interdiff: https://reviewboard.mozilla.org/r/69856/diff/1-2/

This is better, but there are still some regressions on Windows 7 :/ https://screenshots.mattn.ca/compare/?oldProject=mozilla-central&oldRev=6b65dd49d4f045c0a9753ce60bdb4b7b4aaedcf8&newProject=try&newRev=68cd3149f38297c41f70e6186aec80b86e02de67

(In reply to Dão Gottwald [:dao] from comment #5) > This is better, but there are still some regressions on Windows 7 :/ > > https://screenshots.mattn.ca/compare/?oldProject=mozilla- > central&oldRev=6b65dd49d4f045c0a9753ce60bdb4b7b4aaedcf8&newProject=try&newRev > =68cd3149f38297c41f70e6186aec80b86e02de67 How would we narrow down what CSS is causing the regressions? It could be anything under the shared folder, or the Windows specific files. The fact that only Win7 is regressing is kinda weird though.

(In reply to Sumit Tiwari from comment #6) > (In reply to Dão Gottwald [:dao] from comment #5) > > This is better, but there are still some regressions on Windows 7 :/ > > > > https://screenshots.mattn.ca/compare/?oldProject=mozilla- > > central&oldRev=6b65dd49d4f045c0a9753ce60bdb4b7b4aaedcf8&newProject=try&newRev > > =68cd3149f38297c41f70e6186aec80b86e02de67 > > How would we narrow down what CSS is causing the regressions? It could be > anything under the shared folder, or the Windows specific files. The fact > that only Win7 is regressing is kinda weird though. I'm afraid the only way is reviewing the patch manually and making educated guesses, or building locally on Windows and reverting the changes one by one and see what happens :(

(In reply to Dão Gottwald [:dao] from comment #7) > (In reply to Sumit Tiwari from comment #6) > > (In reply to Dão Gottwald [:dao] from comment #5) > > > This is better, but there are still some regressions on Windows 7 :/ > > > > > > https://screenshots.mattn.ca/compare/?oldProject=mozilla- > > > central&oldRev=6b65dd49d4f045c0a9753ce60bdb4b7b4aaedcf8&newProject=try&newRev > > > =68cd3149f38297c41f70e6186aec80b86e02de67 > > > > How would we narrow down what CSS is causing the regressions? It could be > > anything under the shared folder, or the Windows specific files. The fact > > that only Win7 is regressing is kinda weird though. > > I'm afraid the only way is reviewing the patch manually and making educated > guesses, or building locally on Windows and reverting the changes one by one > and see what happens :( I could also push more patches for you to the try server and get screenshots. Getting screenshots takes some time, but we're not in a rush. I could even push a bunch of different experimental patches in parallel.

I'll hopefully get access to a Windows machine soon enough, so I can start experimenting. It would still be helpful to have screenshots though, so that I don't miss out on something.

Is it still actual? I'm from WhatCanIDoForMozilla and BugAhoy.

Hi Paul! Sorry, this bug slipped past the radar - I'm needinfo'ing Dao, the mentor assigned to this bug, for you.

Yeah, this is still relevant. I've been trying to replace #main-window with :root when touching such rules and have pushed for this in reviews as well, so with some luck this bug is bit easier now than it was two years ago. Maybe not a good first bug, though.

I'd like to take this. I'm not on Windows anymore, so I would need somebody to verify it(to not blame all Windows on Nightly users) if this could be still a problem for Windows. Otherwise, I can replace everything to `:root`? I assume I can use artifact builds here?

(In reply to Jens Hausdorf :jens1o [ni? me] from comment #13) > I'd like to take this. I'm not on Windows anymore, so I would need somebody > to verify it(to not blame all Windows on Nightly users) if this could be > still a problem for Windows. We can run the browser-screenshots-e10s suite on the try server to test this on all three platforms. > Otherwise, I can replace everything to `:root`? > I assume I can use artifact builds here? Yes and yes.

Replaced #main-window with :root in the browser/base/content/browser.css

Pushed by dgottwald@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/bacd54533618 Use :root instead of #main-window in browser/base/content/browser.css. r=dao