Bug 206627 – REGRESSION [ Mac wk2 ] Tests are jetsamming: Unable to shrink memory footprint of process below the kill thresold [sic]

• Component: SVG
• Status: RESOLVED FIXED

[Jacob Uphoff]

svg/custom/conditional-processing-2.html is flaky crashing and this started around commit 254816

This crash is happening on Mojave wk2

I did not try to reproduce

History:
https://results.webkit.org/?suite=layout-tests&test=svg%2Fcustom%2Fconditional-processing-2.html

Crasher:

No crash log found for WebKitTestRunner:11765.

stdout:
layer at (0,0) size 800x600
  RenderView at (0,0) size 800x600
layer at (0,0) size 800x162
  RenderBlock {HTML} at (0,0) size 800x162
    RenderBody {BODY} at (8,16) size 784x138
      RenderBlock {P} at (0,0) size 784x18
        RenderText {#text} at (0,0) size 329x18
          text run at (0,0) width 329: "This test passes if four green squares are displayed:"
      RenderBlock (anonymous) at (0,34) size 784x104
        RenderSVGRoot {svg} at (8,50) size 100x100
          RenderSVGForeignObject {foreignObject} at (0,0) size 100x100
            RenderBlock {SPAN} at (0,0) size 100x100 [bgcolor=#008000]
            RenderText {#text} at (0,0) size 0x0
        RenderText {#text} at (100,86) size 4x18
          text run at (100,86) width 4: " "
        RenderSVGRoot {svg} at (112,50) size 100x100
          RenderSVGForeignObject {foreignObject} at (0,0) size 100x100
            RenderMathMLMath {math} at (0,0) size 100x100
              RenderMathMLSpace {mspace} at (0,0) size 100x100 [bgcolor=#008000]
            RenderText {#text} at (0,0) size 0x0
        RenderText {#text} at (204,86) size 4x18
          text run at (204,86) width 4: " "
        RenderSVGRoot {svg} at (216,50) size 100x100
          RenderSVGForeignObject {foreignObject} at (0,0) size 100x100
            RenderBlock {SPAN} at (0,0) size 100x100
              RenderMathMLMath {math} at (0,0) size 100x100
                RenderMathMLSpace {mspace} at (0,0) size 100x100 [bgcolor=#008000]
              RenderText {#text} at (0,0) size 0x0
            RenderText {#text} at (0,0) size 0x0
        RenderText {#text} at (308,86) size 4x18
          text run at (308,86) width 4: " "
        RenderText {#text} at (0,0) size 0x0
        RenderSVGRoot {svg} at (320,50) size 100x100
          RenderSVGRect {rect} at (320,50) size 100x100 [fill={[type=SOLID] [color=#008000]}] [x=0.00] [y=0.00] [width=100.00] [height=100.00]
        RenderText {#text} at (0,0) size 0x0

stderr:
Unable to shrink memory footprint of process (4150 MB) below the kill thresold (4096 MB). Killed
#CRASHED - com.apple.WebKit.WebContent.Development (pid 19092)

[Radar WebKit Bug Importer]

<rdar://problem/58815277>

[Simon Fraser (smfr)]

This only seemed to happen once.

[Alexey Proskuryakov]

Looking just a little bit deeper, no, it's happening all the time, just on different test. This is the most recent test run at the moment:

https://build.webkit.org/builders/Apple-Catalina-Release-WK2-Tests/builds/2521/steps/layout-test/logs/stdio

07:24:09.704 87914 worker/5 svg/custom/inline-style-overrides-markers-expected.svg crashed, (stderr lines):
07:24:09.704 87914   Unable to shrink memory footprint of process (4322 MB) below the kill thresold (4096 MB). Killed
07:24:09.704 87914   #CRASHED - com.apple.WebKit.WebContent.Development (pid 96223)
07:24:09.706 87856 [52288/52814] svg/custom/inline-style-overrides-markers.svg failed unexpectedly (WebKitTestRunner crashed [pid=95002])

[Simon Fraser (smfr)]

Maybe we can dump a memgraph when this happens?

[Ryan Haddad]

*** Bug 206937 has been marked as a duplicate of this bug. ***

[Alexey Proskuryakov]

> Maybe we can dump a memgraph when this happens?

I don't think that we can take a memgraph after the process gets jetsammed.

[Simon Fraser (smfr)]

If ChromeClient implements didExceedBackgroundResourceLimitWhileInForeground, then we IPC the UI process which can kill the web process if it wants to.

So in WTR, we could register this handler (avoiding the instant kill in webcontent), and log something which triggers webkitpy to take a memgraph.

[Ryan Haddad]

Since this jetsam was reproducible by running Web SQL tests, and those tests were being run on the same worker as the blamed SVG tests, I skipped storage/websql tests for iOS and macOS WK2 with https://trac.webkit.org/changeset/255783/webkit