Skip to content

Latest commit

 

History

History
56 lines (44 loc) · 2.36 KB

how-memlab-works.md

File metadata and controls

56 lines (44 loc) · 2.36 KB

How memlab Works

In a nutshell, memlab starts a headless Chrome browser, interacts with the page, takes heap snapshots, and finds memory leaks by parsing, diffing, and analyzing heap snapshots.

1. Browser Interaction

For example, if we want to find memory leaks triggered by some interactions in a web app on a target page (TP), memlab visits the web page in the following order:

  • Visit a different page - let's call it baseline page BP and take a heap snapshot named SBP (the baseline page is specified by the url callback in test scenario)
  • Visit the target page TP and take another heap snapshot STP (the target interactions are specified by the action callback in test scenario)
  • Finally, come back to the baseline page (BP), and take the last heap snapshot SBP'. With these heap snapshots, memlab finds memory leaks as explained in the next section. (the final page is specified by the back callback in test scenario)

2. Heap Analysis

Snapshot decoding: memlab decodes V8 (or hermes) heap snapshots and provides APIs that allows querying JavaScript heap.

Leak detection: A superset of objects leaked from the target page can be derived as follows:

$$(STP \ SBP) ∩ SBP'$$

MemLab first gets a set of allocated objects in TP (target interaction) by excluding SBP's objects (object allocated from the baseline page) from STP (target heap snapshot).

Then it takes an intersection with objects in SBP' (object remaining on the final page) to get objects that:

  1. are allocated from target interaction (TP)
  2. but remain alive after memlab navigates from it.

The built-in leak detectors use domain-specific heuristics to further refine the list of leaked objects (e.g., detached DOM elements, error stack trace, etc. memlab also identifies React Fiber tree and detects unmounted Fiber nodes).

3. Retainer Traces for Memory Leaks

memlab generates retainer traces from GC roots to leaked objects. Sometimes certain interactions could trigger thousands of leaked objects, it would be overwhelming to show all the retainer traces. memlab clusters all retainer traces and only shows one retainer trace for each memory leak cluster.