Bug 237110 – Web Inspector: [Cocoa] Split remote inspector message data into smaller chunks for large messages

• Component: Web Inspector
• Status: RESOLVED FIXED

[Patrick Angle]

<rdar://89364487>

[Patrick Angle]

Created attachment 453054 [details]
Patch v1.0

[Devin Rousso]

Comment on attachment 453054 [details]
Patch v1.0

View in context: https://bugs.webkit.org/attachment.cgi?id=453054&action=review

> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:209
> +    NSUInteger const chunkSize = 2 * 1024 * 1024; // 2 Mebibytes

Style: `const NSUInteger`
NIT: How about `maxChunkSize`?

> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:222
> +        NSUInteger currentChunkSize = (messageLength - offset > chunkSize) ? chunkSize : messageLength - offset;

How about this?
```
NSUInteger currentChunkSize = std::min(messageLength - offset, maxChunkSize);
```
Or if not `std::min` some other similar method?

> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:548
> +    NSNumber *chunkedMessageDataSupportedNumber = userInfo[WIRMessageDataTypeChunkSupportedKey];

NIT: Should we name this (and the member variable) `[m_]messageDataTypeChunkSupported[Number]` instead so that it matches the key?

> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:551
> +    m_chunkedMessageDataSupported = !!chunkedMessageDataSupportedNumber && chunkedMessageDataSupportedNumber.boolValue;

NIT: Is the `!!` needed?  Some of the other `BOOL` seem to just directly consult `.boolValue`, so do we even need the part before the `&&`?

[Devin Rousso]

Comment on attachment 453054 [details]
Patch v1.0

oops, forgot to r=mews

[Joseph Pecoraro]

Comment on attachment 453054 [details]
Patch v1.0

View in context: https://bugs.webkit.org/attachment.cgi?id=453054&action=review

> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:208
> +    NSUInteger messageLength = [messageData length];

I think we'd generally use the property style here `messageData.length`.

> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:221
> +    for (NSUInteger offset = 0; offset < messageLength; offset += chunkSize) {

Would it be worth adding an `@autoreleasepool {}` block inside this loop? I don't know if `subdataWithRange` keeps around chunks of data in 2MB sizes or not. I suspect it doesn't though.

>> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:551
>> +    m_chunkedMessageDataSupported = !!chunkedMessageDataSupportedNumber && chunkedMessageDataSupportedNumber.boolValue;
> 
> NIT: Is the `!!` needed?  Some of the other `BOOL` seem to just directly consult `.boolValue`, so do we even need the part before the `&&`?

I believe technically this line could just be like the one above it:

     m_chunkedMessageDataSupported = chunkedMessageDataSupportedNumber.boolValue;

Because `chunkedMessageDataSupportedNumber.boolValue` is `[chunkedMessageDataSupportedNumber boolValue]` and if `chunkedMessageDataSupportedNumber` is `nil` the result would be `0` which is `NO`.

But I'm fine if you think being more explicit may be more readable.

[Joseph Pecoraro]

Seems some of my comments double posted. Sry about that!

[Patrick Angle]

Comment on attachment 453054 [details]
Patch v1.0

View in context: https://bugs.webkit.org/attachment.cgi?id=453054&action=review

>> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:221
>> +    for (NSUInteger offset = 0; offset < messageLength; offset += chunkSize) {
> 
> Would it be worth adding an `@autoreleasepool {}` block inside this loop? I don't know if `subdataWithRange` keeps around chunks of data in 2MB sizes or not. I suspect it doesn't though.

`subdataWithRange` does not appear to guarantee a no-copy operation, so an `autoreleasepool` is probably a good idea here.

>> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:222
>> +        NSUInteger currentChunkSize = (messageLength - offset > chunkSize) ? chunkSize : messageLength - offset;
> 
> How about this?
> ```
> NSUInteger currentChunkSize = std::min(messageLength - offset, maxChunkSize);
> ```
> Or if not `std::min` some other similar method?

🤦

>>> Source/JavaScriptCore/inspector/remote/cocoa/RemoteInspectorCocoa.mm:551
>>> +    m_chunkedMessageDataSupported = !!chunkedMessageDataSupportedNumber && chunkedMessageDataSupportedNumber.boolValue;
>> 
>> NIT: Is the `!!` needed?  Some of the other `BOOL` seem to just directly consult `.boolValue`, so do we even need the part before the `&&`?
> 
> I believe technically this line could just be like the one above it:
> 
>      m_chunkedMessageDataSupported = chunkedMessageDataSupportedNumber.boolValue;
> 
> Because `chunkedMessageDataSupportedNumber.boolValue` is `[chunkedMessageDataSupportedNumber boolValue]` and if `chunkedMessageDataSupportedNumber` is `nil` the result would be `0` which is `NO`.
> 
> But I'm fine if you think being more explicit may be more readable.

I kinda forgot about the behavior of sending messages to nil objects in ObjC when I wrote this - it can absolutely be reduced down to your suggestions.

[Patrick Angle]

Created attachment 453144 [details]
Patch v1.1 - Address review notes

[EWS]

Committed r290510 (247795@main): <https://commits.webkit.org/247795@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 453144 [details].