Bug 200637 – Web Inspector: REGRESSION: Inspector Timeline always starts at about 500ms mark on a reload

• Status: RESOLVED
• Resolution: FIXED
• Component: Web Inspector

[Joseph Pecoraro]

REGRESSION: Inspector Timeline always starts at about 500ms mark on a reload

[Radar WebKit Bug Importer]

<rdar://problem/54218967>

[Joseph Pecoraro]

I believe this was a regression from r245498, when the CPUTimelineRecord's startTime was moved back by the sampling rate (500ms).

[Joseph Pecoraro]

Created attachment 378616 [details]
[PATCH] Proposed Fix

[Devin Rousso]

Comment on attachment 378616 [details]
[PATCH] Proposed Fix

View in context: https://bugs.webkit.org/attachment.cgi?id=378616&action=review

r-, explained below

> Source/WebInspectorUI/UserInterface/Models/TimelineRecord.js:108
> +    get unadjustedStartTime()

The name of this function doesn't seem accurate, or at the very least seems like a roundabout/circumstantial way of getting to the desired result.  The `timestamp` itself isn't guaranteed to be unadjusted either (I know it is right now, but that may not be the case for other future timeline record subclasses).

I think a better approach would be to check `if (record instanceof WI.CPUTimelineRecord || record instanceof WI.MemoryTimelineRecord)` in `_updateTimesIfNeeded` and use the `record.timestamp` in that case instead.

[Joseph Pecoraro]

Comment on attachment 378616 [details]
[PATCH] Proposed Fix

View in context: https://bugs.webkit.org/attachment.cgi?id=378616&action=review

>> Source/WebInspectorUI/UserInterface/Models/TimelineRecord.js:108
>> +    get unadjustedStartTime()
> 
> The name of this function doesn't seem accurate, or at the very least seems like a roundabout/circumstantial way of getting to the desired result.  The `timestamp` itself isn't guaranteed to be unadjusted either (I know it is right now, but that may not be the case for other future timeline record subclasses).
> 
> I think a better approach would be to check `if (record instanceof WI.CPUTimelineRecord || record instanceof WI.MemoryTimelineRecord)` in `_updateTimesIfNeeded` and use the `record.timestamp` in that case instead.

I'm not sure I buy the argument.

That would mean:
  • The next time we add a timestamp based record we'd have to update that location as well
  • Each update call now has to do a chain of instanceof checks
  • Each TimelineRecord subclass can override unadjusted*Time as needed. So if "timestamp" was adjusted in some future class they would be able to override it to return sane values.

Would you rather I make CPUTimelineRecord / MemoryTimelineRecord provide their own unadjusted*Time (or a common TimestampTimelineRecord superclass) and just return timestamp?

[Devin Rousso]

Comment on attachment 378616 [details]
[PATCH] Proposed Fix

View in context: https://bugs.webkit.org/attachment.cgi?id=378616&action=review

>>> Source/WebInspectorUI/UserInterface/Models/TimelineRecord.js:108
>>> +    get unadjustedStartTime()
>> 
>> The name of this function doesn't seem accurate, or at the very least seems like a roundabout/circumstantial way of getting to the desired result.  The `timestamp` itself isn't guaranteed to be unadjusted either (I know it is right now, but that may not be the case for other future timeline record subclasses).
>> 
>> I think a better approach would be to check `if (record instanceof WI.CPUTimelineRecord || record instanceof WI.MemoryTimelineRecord)` in `_updateTimesIfNeeded` and use the `record.timestamp` in that case instead.
> 
> I'm not sure I buy the argument.
> 
> That would mean:
>   • The next time we add a timestamp based record we'd have to update that location as well
>   • Each update call now has to do a chain of instanceof checks
>   • Each TimelineRecord subclass can override unadjusted*Time as needed. So if "timestamp" was adjusted in some future class they would be able to override it to return sane values.
> 
> Would you rather I make CPUTimelineRecord / MemoryTimelineRecord provide their own unadjusted*Time (or a common TimestampTimelineRecord superclass) and just return timestamp?

Good point.  I like your suggestion.

How about something like:

TimelineRecord.js
```
    get unadjustedStartTime()
    {
        // Overridden by subclasses if needed.
        return this.startTime;
    }
```

CPUTimelineRecord.js
```
    get unadjustedStartTime()
    {
        return this.timestamp;
    }
```

[Joseph Pecoraro]

Created attachment 378620 [details]
[PATCH] Proposed Fix

[Devin Rousso]

Comment on attachment 378620 [details]
[PATCH] Proposed Fix

View in context: https://bugs.webkit.org/attachment.cgi?id=378620&action=review

r=me, thanks for iterating :)

> Source/WebInspectorUI/UserInterface/Models/TimelineRecord.js:108
> +    get unadjustedStartTime()

We should move this to be next to `get activeStartTime`.

> Source/WebInspectorUI/UserInterface/Models/TimelineRecord.js:114
> +    get unadjustedEndTime()

We should move this to be right after `get endTime`.

[Joseph Pecoraro]

https://trac.webkit.org/changeset/249799/webkit