211013 2020-04-25 00:24:52 -0700 getShaderPrecisionFormat returns the wrong values 2020-05-02 21:55:34 -0700 1 1 1 Unclassified WebKit WebGL WebKit Nightly Build All All RESOLVED FIXED InRadar P2 Normal --- 1 gman dino darin dino ews-watchlist graouts kbr kondapallykalyan webkit-bug-importer oldest_to_newest 1645549 0 gman 2020-04-25 00:24:52 -0700 Checking the code it's appears to be hard coded https://trac.webkit.org/browser/webkit/trunk/Source/WebCore/platform/graphics/opengl/GraphicsContextGLOpenGLBase.cpp#L362 1645701 1 webkit-bug-importer 2020-04-25 17:27:45 -0700 <rdar://problem/62378056> 1645863 2 397638 dino 2020-04-26 17:42:13 -0700 Created attachment 397638 Patch 1645866 3 397638 darin 2020-04-26 17:58:47 -0700 Comment on attachment 397638 Patch This wasn’t an anti-fingerprinting measure? 1646107 4 dino 2020-04-27 11:30:54 -0700 (In reply to Darin Adler from comment #3) > Comment on attachment 397638 [details] > Patch > > This wasn’t an anti-fingerprinting measure? That's a good point. Let me look at the history. 1646124 5 dino 2020-04-27 12:11:03 -0700 (In reply to Darin Adler from comment #3) > Comment on attachment 397638 [details] > Patch > > This wasn’t an anti-fingerprinting measure? They were always exposed as constant values! https://trac.webkit.org/changeset/113092/webkit + // These constants came from the Chromium port; we believe they originally + // came from making the actual API call on a representative desktop system. Unfortunately my macOS systems return the same value, so I can't see if Chrome/Edge + Firefox do this properly. I'll have to look at their code. For iOS, this does expose different values (from macOS). I'll also check an Android device. 1646170 6 kbr 2020-04-27 13:30:29 -0700 The original code came from early WebGL implementations, where this API call only existed on mobile devices and not desktop. It's important for correctness and portability to expose this accurately. Gregg's encountered portability issues and a new WebGL conformance test which fails without this change is incoming in https://github.com/KhronosGroup/WebGL/pull/3061 . 1646173 7 dino 2020-04-27 13:34:18 -0700 The main issue is that the hard-coded values are higher than what we support on iOS, which could lead to people thinking the device is more powerful (or accurate) than it really is. So if we're going to hard-code values, we should pick the lowest. 1646174 8 dino 2020-04-27 13:36:50 -0700 (In reply to Kenneth Russell from comment #6) > The original code came from early WebGL implementations, where this API call > only existed on mobile devices and not desktop. It's strange then that we hard-coded it to a desktop value then, and never asked for the real value on a mobile device. > > It's important for correctness and portability to expose this accurately. > Gregg's encountered portability issues and a new WebGL conformance test > which fails without this change is incoming in > https://github.com/KhronosGroup/WebGL/pull/3061 . It does expose a fingerprint though. I think I need to check how much the values can vary between devices (desktop to desktop, and mobile to mobile). 1646196 9 kbr 2020-04-27 14:06:00 -0700 This will likely mainly allow distinguishing between desktop and mobile GPUs, especially since there are relatively few GPU types on both macOS and iOS. The user agent string already provides essentially this information, in allowing distinguishing between operating systems. 1646256 10 dino 2020-04-27 15:04:57 -0700 Apple's OpenGL team confirmed that all iOS devices return the same values, as do all macOS devices. So this isn't extra fingerprinting. 1646258 11 dino 2020-04-27 15:08:44 -0700 Committed r260789: <https://trac.webkit.org/changeset/260789> 1648311 12 gman 2020-05-02 21:55:34 -0700 Just FYI. Even if they were different on every GPU not exposing them would not add new fingerprinting. The whole point of the new conformance test is to actually render to find out what the precision is because the value was wrong here. I'm sure a better test could derive any precision even if it was not reported here 397638 2020-04-26 17:42:13 -0700 2020-04-26 17:58:25 -0700 Patch bug-211013-20200427104212.patch text/plain 2827 dino U3VidmVyc2lvbiBSZXZpc2lvbjogMjYwNjY3CmRpZmYgLS1naXQgYS9Tb3VyY2UvV2ViQ29yZS9D aGFuZ2VMb2cgYi9Tb3VyY2UvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXggOTdlMTAzNDgwODM5NmRh YjAwZWUxZDIwOTYxNDI1YjZjYTk0ODllNy4uZmQxNjQ5OGQ4ZmE0YTQxNGEzZjFiY2NlZmNmM2E5 ZDU4MTMxNDI3MyAxMDA2NDQKLS0tIGEvU291cmNlL1dlYkNvcmUvQ2hhbmdlTG9nCisrKyBiL1Nv dXJjZS9XZWJDb3JlL0NoYW5nZUxvZwpAQCAtMSwzICsxLDIzIEBACisyMDIwLTA0LTI2ICBEZWFu IEphY2tzb24gIDxkaW5vQGFwcGxlLmNvbT4KKworICAgICAgICBnZXRTaGFkZXJQcmVjaXNpb25G b3JtYXQgcmV0dXJucyB0aGUgd3JvbmcgdmFsdWVzCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJr aXQub3JnL3Nob3dfYnVnLmNnaT9pZD0yMTEwMTMKKyAgICAgICAgPHJkYXI6Ly9wcm9ibGVtLzYy Mzc4MDU2PgorCisgICAgICAgIFJldmlld2VkIGJ5IE5PQk9EWSAoT09QUyEpLgorCisgICAgICAg IEdyZWdnIHBvaW50ZWQgb3V0IHRoYXQgb3VyIGNvZGUgaGFyZGNvZGVzIHZhbHVlcyBmb3IKKyAg ICAgICAgZ2V0U2hhZGVyUHJlY2lzaW9uRm9ybWF0LiBUaGUgZml4IGlzIHNpbXBseSB0byBjYWxs IGludG8KKyAgICAgICAgdGhlIHVuZGVybHlpbmcgR0wgZnVuY3Rpb24uCisKKyAgICAgICAgQ292 ZXJlZCBieSB0aGUgZXhpc3RpbmcgdGVzdDogd2ViZ2wvMS4wLjMvY29uZm9ybWFuY2UvbWlzYy9z aGFkZXItcHJlY2lzaW9uLWZvcm1hdC5odG1sCisKKyAgICAgICAgSG93ZXZlciwgdGhhdCBqdXN0 IHRlc3RzIG1pbmltdW0gdmFsdWVzLiBEZXZpY2VzIGhhdmUgZGlmZmVyZW50CisgICAgICAgIGFj dHVhbCByZXN1bHRzLgorCisgICAgICAgICogcGxhdGZvcm0vZ3JhcGhpY3MvYW5nbGUvR3JhcGhp Y3NDb250ZXh0R0xBTkdMRS5jcHA6CisgICAgICAgIChXZWJDb3JlOjpHcmFwaGljc0NvbnRleHRH TE9wZW5HTDo6Z2V0U2hhZGVyUHJlY2lzaW9uRm9ybWF0KToKKwogMjAyMC0wNC0yNCAgVGltIEhv cnRvbiAgPHRpbW90aHlfaG9ydG9uQGFwcGxlLmNvbT4KIAogICAgICAgICBpUGFkOiAiUG9ja2V0 IENpdHkiIGludGVyYWN0aW9uIGRvZXMgbm90IHdvcmsgd2l0aCB0cmFja3BhZApkaWZmIC0tZ2l0 IGEvU291cmNlL1dlYkNvcmUvcGxhdGZvcm0vZ3JhcGhpY3MvYW5nbGUvR3JhcGhpY3NDb250ZXh0 R0xBTkdMRS5jcHAgYi9Tb3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGljcy9hbmdsZS9HcmFw aGljc0NvbnRleHRHTEFOR0xFLmNwcAppbmRleCA1YjA1NmVkMjY3ZjMzY2MxY2IwNDU1MWVkNzFm ZDQ4NjNiNjk4Y2VmLi4wZWQ1YTE0MTdmZWU4NjlmMTM5MDRlMjRlNTQzMTYxZGNlNWE5ZDU3IDEw MDY0NAotLS0gYS9Tb3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGljcy9hbmdsZS9HcmFwaGlj c0NvbnRleHRHTEFOR0xFLmNwcAorKysgYi9Tb3VyY2UvV2ViQ29yZS9wbGF0Zm9ybS9ncmFwaGlj cy9hbmdsZS9HcmFwaGljc0NvbnRleHRHTEFOR0xFLmNwcApAQCAtMzExLDMzICszMTEsOCBAQCB2 b2lkIEdyYXBoaWNzQ29udGV4dEdMT3BlbkdMOjpnZXRJbnRlZ2VydihHQ0dMZW51bSBwbmFtZSwg R0NHTGludCogdmFsdWUpCiAKIHZvaWQgR3JhcGhpY3NDb250ZXh0R0xPcGVuR0w6OmdldFNoYWRl clByZWNpc2lvbkZvcm1hdChHQ0dMZW51bSBzaGFkZXJUeXBlLCBHQ0dMZW51bSBwcmVjaXNpb25U eXBlLCBHQ0dMaW50KiByYW5nZSwgR0NHTGludCogcHJlY2lzaW9uKQogewotICAgIFVOVVNFRF9Q QVJBTShzaGFkZXJUeXBlKTsKLSAgICBBU1NFUlQocmFuZ2UpOwotICAgIEFTU0VSVChwcmVjaXNp b24pOwotCiAgICAgbWFrZUNvbnRleHRDdXJyZW50KCk7Ci0KLSAgICBzd2l0Y2ggKHByZWNpc2lv blR5cGUpIHsKLSAgICBjYXNlIEdyYXBoaWNzQ29udGV4dEdMOjpMT1dfSU5UOgotICAgIGNhc2Ug R3JhcGhpY3NDb250ZXh0R0w6Ok1FRElVTV9JTlQ6Ci0gICAgY2FzZSBHcmFwaGljc0NvbnRleHRH TDo6SElHSF9JTlQ6Ci0gICAgICAgIC8vIFRoZXNlIHZhbHVlcyBhcmUgZm9yIGEgMzItYml0IHR3 b3MtY29tcGxlbWVudCBpbnRlZ2VyIGZvcm1hdC4KLSAgICAgICAgcmFuZ2VbMF0gPSAzMTsKLSAg ICAgICAgcmFuZ2VbMV0gPSAzMDsKLSAgICAgICAgcHJlY2lzaW9uWzBdID0gMDsKLSAgICAgICAg YnJlYWs7Ci0gICAgY2FzZSBHcmFwaGljc0NvbnRleHRHTDo6TE9XX0ZMT0FUOgotICAgIGNhc2Ug R3JhcGhpY3NDb250ZXh0R0w6Ok1FRElVTV9GTE9BVDoKLSAgICBjYXNlIEdyYXBoaWNzQ29udGV4 dEdMOjpISUdIX0ZMT0FUOgotICAgICAgICAvLyBUaGVzZSB2YWx1ZXMgYXJlIGZvciBhbiBJRUVF IHNpbmdsZS1wcmVjaXNpb24gZmxvYXRpbmctcG9pbnQgZm9ybWF0LgotICAgICAgICByYW5nZVsw XSA9IDEyNzsKLSAgICAgICAgcmFuZ2VbMV0gPSAxMjc7Ci0gICAgICAgIHByZWNpc2lvblswXSA9 IDIzOwotICAgICAgICBicmVhazsKLSAgICBkZWZhdWx0OgotICAgICAgICBBU1NFUlRfTk9UX1JF QUNIRUQoKTsKLSAgICAgICAgYnJlYWs7Ci0gICAgfQorICAgIGdsOjpHZXRTaGFkZXJQcmVjaXNp b25Gb3JtYXQoc2hhZGVyVHlwZSwgcHJlY2lzaW9uVHlwZSwgcmFuZ2UsIHByZWNpc2lvbik7CiB9 CiAKIGJvb2wgR3JhcGhpY3NDb250ZXh0R0xPcGVuR0w6OnRleEltYWdlMkQoR0NHTGVudW0gdGFy Z2V0LCBHQ0dMaW50IGxldmVsLCBHQ0dMZW51bSBpbnRlcm5hbGZvcm1hdCwgR0NHTHNpemVpIHdp ZHRoLCBHQ0dMc2l6ZWkgaGVpZ2h0LCBHQ0dMaW50IGJvcmRlciwgR0NHTGVudW0gZm9ybWF0LCBH Q0dMZW51bSB0eXBlLCBjb25zdCB2b2lkKiBwaXhlbHMpCg==