Bug 193262

Summary: Web Inspector: expose Audit and Recording versions to the frontend
Product: WebKit Reporter: Devin Rousso <hi>
Component: Web InspectorAssignee: Devin Rousso <hi>
Status: RESOLVED FIXED    
Severity: Normal CC: bburg, commit-queue, ews-watchlist, hi, inspector-bugzilla-changes, joepeck, keith_miller, mark.lam, msaboff, saam, webkit-bug-importer
Priority: P2 Keywords: InRadar
Version: WebKit Nightly Build   
Hardware: All   
OS: All   
Bug Depends on: 190754, 193149    
Bug Blocks: 173807, 193686    
Attachments:
Description Flags
Patch
none
[Image] After Patch is applied
none
[Patch] Command Version
none
[Patch] Protocol Version
none
[Patch] Protocol Version
none
[Patch] Protocol Version
none
Patch none

Devin Rousso
Reported 2019-01-08 15:43:26 PST
In order to have versioning support for Audits, we need a place where we can show version information (e.g. "Recording System Version 1"). These version values are split into two "types": - Interface: the WebInspector frontend's ability to display the data (e.g. whether audits can show metadata information) - System: the inspected target's ability to generate/process/execute the data (e.g. whether audits are able to use the injected `AUDIT` object)
Attachments
Patch (25.82 KB, patch)
2019-01-08 16:21 PST, Devin Rousso 		no flags
DetailsFormatted DiffDiff
[Image] After Patch is applied (636.32 KB, image/png)
2019-01-08 16:22 PST, Devin Rousso 		no flags
Details
[Patch] Command Version (18.82 KB, patch)
2019-01-12 20:10 PST, Devin Rousso 		no flags
DetailsFormatted DiffDiff
[Patch] Protocol Version (72.71 KB, patch)
2019-01-15 20:35 PST, Devin Rousso 		no flags
DetailsFormatted DiffDiff
[Patch] Protocol Version (71.96 KB, patch)
2019-01-15 21:00 PST, Devin Rousso 		no flags
DetailsFormatted DiffDiff
[Patch] Protocol Version (72.16 KB, patch)
2019-01-22 15:40 PST, Devin Rousso 		no flags
DetailsFormatted DiffDiff
Patch (70.90 KB, patch)
2019-01-22 19:38 PST, Devin Rousso 		no flags
DetailsFormatted DiffDiff
Show Obsolete (6) View All Add attachment proposed patch, testcase, etc.
Radar WebKit Bug Importer
Comment 1 2019-01-08 15:44:27 PST
<rdar://problem/47130684>
Devin Rousso
Comment 2 2019-01-08 16:21:47 PST
Created attachment 358648 [details] Patch Not sure if "Interface" and "System" are really the right `UIString`s for this, but I figured I'd put the code up and then worry about it instead of the other way around 😅
Devin Rousso
Comment 3 2019-01-08 16:22:00 PST
Created attachment 358649 [details] [Image] After Patch is applied
Devin Rousso
Comment 4 2019-01-12 20:10:39 PST
Created attachment 359000 [details] [Patch] Command Version Removed the UI parts of this patch to focus more on the functionality. Added a requirement that the system version be kept in sync with the frontend version (enforced by a test) so that when doing a "compatibility" check we only need to do compare the listed version against the minimum of the system and frontend versions: - if system < version < frontend (newer frontend connected to older backend), then we lack required functionality - if frontend < version < system (older frontend connected to newer backend), then we are able to run the audit, but may be unable to display certain result values - if version < system <=> frontend, the test is fully supported - if system <=> frontend < version, the test is not supported
Joseph Pecoraro
Comment 5 2019-01-14 19:34:39 PST
Comment on attachment 359000 [details] [Patch] Command Version View in context: https://bugs.webkit.org/attachment.cgi?id=359000&action=review r=me > Source/JavaScriptCore/inspector/protocol/Audit.json:9 > + "types": [ > + { > + "id": "Version", > + "type": "number" > + } > + ], This should give this a description. For example in this case the importance behind the Version number is that the tests that are run have access to a global `WebInspectorAudit` API object. Were that to change this Version would need to change to indicate a new type of API surface. Side note: It might even be nice to eventually turn this into a constant right in the protocol file. { id: "Version", type: "number", value: 1, description: "..." }. The tricky bit would be $ref: "Version" would need to know to fill the value at generation time. > Source/JavaScriptCore/inspector/protocol/Recording.json:9 > + { > + "id": "Version", > + "type": "number" > + }, This should also use a description. The importance of this version number is that the backend generates the JSON recording format, and if the JSON needs to change in any material way the version number would need to change. > Source/WebInspectorUI/UserInterface/Models/Recording.js:55 > - if (isNaN(payload.version) || payload.version <= 0) > + if (isNaN(payload.version) || payload.version <= 0 || payload.version > WI.Recording.Version) > return null; Maybe this should include a warning reason dumped to the console. I think there is already a warning dumped to the console when an import fails, but knowing it is because the version number might be useful. Nobody will run into this if the version doesn't change, but when it does...
Devin Rousso
Comment 6 2019-01-15 20:31:49 PST
Comment on attachment 359000 [details] [Patch] Command Version View in context: https://bugs.webkit.org/attachment.cgi?id=359000&action=review >> Source/JavaScriptCore/inspector/protocol/Audit.json:9 >> + ], > > This should give this a description. For example in this case the importance behind the Version number is that the tests that are run have access to a global `WebInspectorAudit` API object. Were that to change this Version would need to change to indicate a new type of API surface. > > Side note: It might even be nice to eventually turn this into a constant right in the protocol file. { id: "Version", type: "number", value: 1, description: "..." }. The tricky bit would be $ref: "Version" would need to know to fill the value at generation time. I was actually initially playing around with adding a constant to the protocol, but I was more focused on building a UI at that point. >> Source/WebInspectorUI/UserInterface/Models/Recording.js:55 >> return null; > > Maybe this should include a warning reason dumped to the console. I think there is already a warning dumped to the console when an import fails, but knowing it is because the version number might be useful. Nobody will run into this if the version doesn't change, but when it does... The default error is "unsupported version", so it should cover this. We should make the error messages a bit better (more comprehensive). The same is true for the various audit model objects (default error is "invalid JSON").
Devin Rousso
Comment 7 2019-01-15 20:32:23 PST
Comment on attachment 359000 [details] [Patch] Command Version View in context: https://bugs.webkit.org/attachment.cgi?id=359000&action=review >>> Source/WebInspectorUI/UserInterface/Models/Recording.js:55 >>> return null; >> >> Maybe this should include a warning reason dumped to the console. I think there is already a warning dumped to the console when an import fails, but knowing it is because the version number might be useful. Nobody will run into this if the version doesn't change, but when it does... > > The default error is "unsupported version", so it should cover this. We should make the error messages a bit better (more comprehensive). > > The same is true for the various audit model objects (default error is "invalid JSON"). <https://webkit.org/b/193476>
Devin Rousso
Comment 8 2019-01-15 20:35:45 PST
Created attachment 359252 [details] [Patch] Protocol Version
EWS Watchlist
Comment 9 2019-01-15 20:36:42 PST Comment hidden (obsolete)
EWS Watchlist
Comment 10 2019-01-15 20:36:55 PST Comment hidden (obsolete)
Devin Rousso
Comment 11 2019-01-15 21:00:04 PST
Created attachment 359254 [details] [Patch] Protocol Version
EWS Watchlist
Comment 12 2019-01-15 21:03:31 PST Comment hidden (obsolete)
Joseph Pecoraro
Comment 13 2019-01-22 14:52:52 PST
Comment on attachment 359254 [details] [Patch] Protocol Version View in context: https://bugs.webkit.org/attachment.cgi?id=359254&action=review r- because I'd like to see another version of this patch. I'd suggest: • Limiting to just `int` • In the frontend making it `Agent.version`. • Potential codegen error if a command is named `version` given it would conflict with `Agent.version`, but this is not important as I don't expect this to ever happen. > Source/WebInspectorUI/ChangeLog:12 > + Add "Interface" version values. Why are there quotes around Interface? > Source/JavaScriptCore/inspector/scripts/codegen/generate_cpp_protocol_types_header.py:107 > + if isinstance(version, (int, long, float)): > + domain_lines.append('static const auto VERSION = %s;' % version) > + elif isinstance(version, (str, unicode)): > + domain_lines.append('static const auto VERSION = "%s";' % version) Lets use explicit types. `int` or `unsigned` for the version number and `const char*`. > Source/JavaScriptCore/inspector/scripts/codegen/generate_cpp_protocol_types_header.py:114 > + else: Style: Drop the else > Source/JavaScriptCore/inspector/scripts/codegen/generate_js_backend_commands.py:91 > + if isinstance(version, (int, long, float)): > + lines.append('InspectorBackend.registerVersion("%(domain)s", %(version)s);' % version_args) > + elif isinstance(version, (str, unicode)): > + lines.append('InspectorBackend.registerVersion("%(domain)s", "%(version)s");' % version_args) This doesn't appear in the code-get test result of InspectorBackendCommands.js: Source/JavaScriptCore/inspector/scripts/tests/generic/expected/version.json-result Looks like you will need to update `should_generate_domain` to say YES if there is a version. > Source/JavaScriptCore/inspector/scripts/codegen/models.py:378 > + if not isinstance(json['version'], (int, float, str, unicode)): > + raise ParseException("Malformed domain specification: version is not a number or string") You allowed `long` in other places but only `int` here. Why not just standardize on int? > Source/WebInspectorUI/UserInterface/Protocol/InspectorBackend.js:145 > + var agent = this._agentForDomain(domainName); Style: `let`. > Source/WebInspectorUI/UserInterface/Protocol/InspectorBackend.js:146 > + agent.VERSION = version; Why not `agent.version`? The capital letters do not add anything. > Source/WebInspectorUI/UserInterface/Protocol/InspectorBackend.js:249 > + this._version = undefined; This is unused.
Joseph Pecoraro
Comment 14 2019-01-22 14:54:10 PST
Comment on attachment 359254 [details] [Patch] Protocol Version View in context: https://bugs.webkit.org/attachment.cgi?id=359254&action=review >> Source/JavaScriptCore/inspector/scripts/codegen/generate_js_backend_commands.py:91 >> + lines.append('InspectorBackend.registerVersion("%(domain)s", "%(version)s");' % version_args) > > This doesn't appear in the code-get test result of InspectorBackendCommands.js: > Source/JavaScriptCore/inspector/scripts/tests/generic/expected/version.json-result > > Looks like you will need to update `should_generate_domain` to say YES if there is a version. Err, I meant "in the codegen test result"
Devin Rousso
Comment 15 2019-01-22 15:40:26 PST
Created attachment 359790 [details] [Patch] Protocol Version
EWS Watchlist
Comment 16 2019-01-22 15:43:27 PST Comment hidden (obsolete)
Joseph Pecoraro
Comment 17 2019-01-22 17:00:59 PST
Comment on attachment 359790 [details] [Patch] Protocol Version View in context: https://bugs.webkit.org/attachment.cgi?id=359790&action=review r=me > Source/WebInspectorUI/UserInterface/Protocol/Connection.js:306 > + if (agent.VERSION !== undefined) > + clone.VERSION = agent.VERSION; Is this necessary? var foo = {version:1}; var bar = Object.create(foo); assert(bar.version === 1); // true > Source/WebInspectorUI/UserInterface/Protocol/Connection.js:331 > + if (agent.VERSION !== undefined) > + clone.VERSION = agent.VERSION; Ditto.
Devin Rousso
Comment 18 2019-01-22 19:38:26 PST
Created attachment 359837 [details] Patch
EWS Watchlist
Comment 19 2019-01-22 19:40:10 PST Comment hidden (obsolete)
WebKit Commit Bot
Comment 20 2019-01-22 20:17:12 PST
Comment on attachment 359837 [details] Patch Clearing flags on attachment: 359837 Committed r240318: <https://trac.webkit.org/changeset/240318>
WebKit Commit Bot
Comment 21 2019-01-22 20:17:14 PST
All reviewed patches have been landed. Closing bug.
Note You need to log in before you can comment on or make changes to this bug.