-
Notifications
You must be signed in to change notification settings - Fork 378
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Remove serialization workarounds for ie 6/7 and rhino (#9578) #9876
base: main
Are you sure you want to change the base?
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for taking this on, legacy code can be a slog to remove... I will note for posterity that at least up to IE9 (and I think 10?) there is some memory leak that occurs in this code as well, but I can't remember the specifics - but since even IE11 is effectively dead, we should go ahead and ignore this.
It looks like this still leaves the eval()
call in the client code, if the server ends up sending version < 8
(and the default is 7), see
Lines 61 to 64 in 0654587
if (readVersion(encoded) < SERIALIZATION_STREAM_JSON_VERSION) { | |
// Versions prior to 8 uses almost JSON with Javascript is special cases; e.g., using ].concat([, | |
// non-stringified NaN/Infinity or ''+'' concatenated strings. | |
results = eval(encoded); |
gwt/user/src/com/google/gwt/user/client/rpc/impl/AbstractSerializationStream.java
Line 50 in 2b23c5e
public static final int SERIALIZATION_STREAM_VERSION = 7; |
I can see an argument for removing this in phases, so that old clients will still work - but if that's the plan, I'd want to consider bumping the SERIALIZATION_STREAM_VERSION to 8, and warning on old versions.
Also: the rhino notes are about the legacy dev mode implementation of gwt-rpc (to avoid needing to emulate all the string operations in JS). It is technically possible that this will break GWTTestCase for htmlunit "hosted mode" tests (hosted mode == legacy dev mode)
gwt/user/src/com/google/gwt/user/client/rpc/impl/ClientSerializationStreamReader.java
Lines 70 to 75 in 2b23c5e
* The payload is almost a JSON literal except for some nuances, like unicode and array concats. | |
* We have a specialized devmode version to decode this payload, because the webmode version | |
* requires multiple round-trips between devmode and the JSVM for every single element in the | |
* payload. This can require several seconds to decode a single RPC payload. The RpcDecoder | |
* operates by doing a limited JS parse on the payload within the devmode VM using Rhino, | |
* avoiding the cross-process RPCs. |
To deal with that and still kill this old code, I'd consider removing the linked file above and replacing it with the supersource version (the second link in this reply, except with the @GwtScriptOnly
removed, and javadoc updated to indicate that it is no longer supersourced).
What do you think?
writePayload(stream); | ||
writeStringTable(stream); | ||
writeHeader(stream); | ||
StringBuffer buffer = new StringBuffer(capacityGuess); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
As long as we're updating this, consider replacing StringBuffer with StringBuilder for better single-threaded performance? From https://docs.oracle.com/en/java/javase/17/docs/api/java.base/java/lang/StringBuffer.html
As of release JDK 5, this class has been supplemented with an equivalent class designed for use by a single thread, StringBuilder. The StringBuilder class should generally be used in preference to this one, as it supports all of the same operations but it is faster, as it performs no synchronization.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I updated to use StringBuilder instead of StringBuffer and changed the default serialization format to JSON.
I'm not sure how best to implement the rest of your suggestions though. But they seem logical.
: Integer.MAX_VALUE; | ||
|
||
while (i < length && charVector.getSize() < maxSegmentVectorSize) { | ||
while (i < length) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This can be removed. The same loop definition is in the line above.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Removed the extra loop
Any chance to get this PR in the next release (because this enabled a CSP w/o "unsafe-eval")? |
Anyone who wants to pick this up and continue it, removing other related dead code, would be welcome to do so. I don't presently have a plan to take this on, so we're looking for a contributor or sponsor to focus on it. See notes in my review above for other steps that should be taken to correct this more than superficially. |
Would it be possible to just merge this now and make a separate issue for the dead code removal? The current code fails with any site that uses CSP under the specific conditions of the issue as |
@codemasterover9000 to confirm, the work you're doing here mostly checks two boxes so far:
Unless I'm mistaken, you could just configure your own project to use version 8, and avoid the codepaths that you are removing in this PR? That would solve this except for too-big payloads (strings longer than MAX_STRING_NODE_LENGTH, - a much smaller PR could update that with no risk of being half finished, and just make that constant able to be configured by another system property, so you could set the limit as high as you wanted (Integer.MAX_VALUE for example). I'm guessing you haven't run all tests on this PR, but looking briefly at some dev mode tests (which use rhino in legacy dev mode) it appears that some tests may fail from this change. Notably, here's the test that validates that Rhino still has the bug (and so this PR will break it): gwt/user/test/com/google/gwt/user/client/rpc/impl/ClientSerializationStreamReaderTest.java Lines 76 to 100 in 89d06d6
Running the
I show one failure: ValueTypesTest.testString_over64KBWithUnicode()
That does not fail before your changes. I have not yet run the full set of tests, was waiting for the branch to be more completed to try to proceed. What do you think of trying to limit the scope of the patch as I proposed above, so that it can be disabled on a per-project basis, until we manage these remaining issues and complete the related work? |
Yes we use version 8. I agree that making the threshold a system property as you suggested would be a better option for a less impactful change. I made a new pull request here #9961 |
Removed the workarounds for IE6/7 issues and rhino. IE 6/7 are not relevant anymore today. Not sure if this is the rhino issue referenced in the code: https://bugzilla.mozilla.org/show_bug.cgi?id=179068. But it seems to have been fixed a long time ago.
Fixes #9578