Repository navigation
Core: Throws an error if the document element is null - #439
Conversation
|
|
| // Return early if doc is invalid or already selected | ||
| if ( doc === document || doc.nodeType !== 9 || !doc.documentElement ) { | ||
| if ( !document ) { | ||
| throw new Error( "Sizzle requires the document element" ); |
There was a problem hiding this comment.
I don't want to start synthesizing errors for bad environment, but I would be willing to remove some of these guards and let errors happen naturally. Please try updating the early return condition to just if ( doc === document ) and see what happens with the test suite.
There was a problem hiding this comment.
-
if ( doc === document )- couldn't build, got errorsPhantomJS 2.1.1 (Linux 0.0.0) selector class FAILED Died on test #24 global code@test/unit/selector.js:256:11: undefined is not a constructor (evaluating 'document.createElement("fieldset")') assert@dist/sizzle.js:387:33 setDocument@dist/sizzle.js:612:29 matchesSelector@dist/sizzle.js:988:14 test/unit/selector.js:298:36 runTest@node_modules/qunitjs/qunit/qunit.js:843:32 run@node_modules/qunitjs/qunit/qunit.js:828:11 node_modules/qunitjs/qunit/qunit.js:970:14 process@node_modules/qunitjs/qunit/qunit.js:629:24 begin@node_modules/qunitjs/qunit/qunit.js:611:9 node_modules/qunitjs/qunit/qunit.js:671:9 Expected 27 assertions, but 24 were run global code@test/unit/selector.js:256:11 ... -
if ( doc === document || doc.nodeType !== 9 )- built, tested, no errorsThough it's pretty much like the original version, it produces a better error
TypeError: "docElem is null"ifdocument.documentElement === null, instead ofTypeError: "document is undefined"
This is not related to this PR: test/index.html every test fails on Firefox (beforeEach failed on element: context is null), Chrome wouldn't load the test (jquery.js:8526 Access to XMLHttpRequest at 'Sizzle/test/data/fixtures.html' from origin 'null' has been blocked by CORS policy: Cross origin requests are only supported for protocol schemes: http, data, chrome, chrome-extension, https.) but it works on Edge - all the tests pass, is this a known issue or should I open a new issue?
Closes jquerygh-439 (cherry picked from commit 7d92424)
This is a resubmit of jquerygh-439 that was lost during the switch from JSHint + JSCS to ESLint Ref jquerygh-442 Ref jquerygh-439 (cherry picked from commit 7d92424)
With new selector code doing less convoluted support tests, it was possible to extract a lot of logic out of setDocument & also reduce size. This commit also backports jquery/sizzle#439 that was reverted by mistake during a switch from JSHint + JSCS to ESLint. Ref jquery/sizzle#442 Ref jquery/sizzle#439
With new selector code doing less convoluted support tests, it was possible to extract a lot of logic out of setDocument & also reduce size. This commit also backports jquery/sizzle#439 that was reverted by mistake during a switch from JSHint + JSCS to ESLint. Ref jquery/sizzle#442 Ref jquery/sizzle#439
With new selector code doing less convoluted support tests, it was possible to extract a lot of logic out of setDocument & also reduce size. This commit also backports jquery/sizzle#439 that was reverted by mistake during a switch from JSHint + JSCS to ESLint. Ref jquery/sizzle#442 Ref jquery/sizzle#439
With new selector code doing less convoluted support tests, it was possible to extract a lot of logic out of setDocument & also reduce size. This commit also backports jquery/sizzle#439 that was reverted by mistake during a switch from JSHint + JSCS to ESLint. Ref jquery/sizzle#442 Ref jquery/sizzle#439
This is a resubmit of jquerygh-439 that was lost during the switch from JSHint + JSCS to ESLint Ref jquerygh-442 Ref jquerygh-439 (cherry picked from commit 7d92424)
With new selector code doing less convoluted support tests, it was possible to extract a lot of logic out of setDocument & also reduce size. This commit also backports jquery/sizzle#439 that was reverted by mistake during a switch from JSHint + JSCS to ESLint. Ref jquery/sizzle#442 Ref jquery/sizzle#439
With new selector code doing less convoluted support tests, it was possible to extract a lot of logic out of setDocument & also reduce size. This commit also backports jquery/sizzle#439 that was reverted by mistake during a switch from JSHint + JSCS to ESLint. Ref jquery/sizzle#442 Ref jquery/sizzle#439
With new selector code doing less convoluted support tests, it was possible to extract a lot of logic out of setDocument & also reduce size. This commit also backports jquery/sizzle#439 that was reverted by mistake during a switch from JSHint + JSCS to ESLint. Closes gh-4462 Ref jquery/sizzle#442 Ref jquery/sizzle#439
|
For the record, this change was accidentally reverted in #442 (62061d0) and I tried to bring it back in #452 but I discovered it breaks running Sizzle with context inside of a I created #458 to not lose track of an issue. That said, it's not likely we'll fix the issue unless someone submits a PR as Sizzle is slowly getting retired and jQuery 4.0 will no longer depend on it. |
4 participants
If Sizzle/jQuery is injected to a page before the document element is constructed (
window.document.documentElement === null), it fails to initialize. I would like to reference jquery/jquery#3333 to describe the issue.setDocument()however simply returnsdocumentin such case -- even if the variabledocumenthasn't been assigned yet (when Sizzle is initializing).It later produces an error
TypeError: "document is undefined"inassert()though the problem was actually there whensetDocument()silently returned. This makes it harder for a dev-user to know what's just happened aswindow.documentexists butwindow.document.documentElementdoesn't. This pull request is intended to make the error verbose.