1401224 - Assertion failure: aLength > 0 (Writing a zero-sized chunk), at image/SourceBuffer.cpp:348

Status: RESOLVED FIXED

(Core :: Networking: Cache, defect)

Description

[Bob Clary [:bc] (inactive)]

Attachment: Windows Debug Log

1. https://nhiemvu.lienminh.garena.vn/tanthu
2. Assertion failure: aLength > 0 (Writing a zero-sized chunk), at z:/build/build/src/image/SourceBuffer.cpp:348
#01: mozilla::image::SourceBuffer::Append(char const *,unsigned int) [image/SourceBuffer.cpp:347]
#02: mozilla::image::AppendToSourceBuffer [image/SourceBuffer.cpp:454]
#03: mozilla::net::CacheFileInputStream::ReadSegments(nsresult (*)(nsIInputStream *,void *,char const *,unsigned int,unsigned int,unsigned int *),void *,unsigned int,unsigned int *) [netwerk/cache2/CacheFileInputStream.cpp:186]
#04: mozilla::image::SourceBuffer::AppendFromInputStream(nsIInputStream *,unsigned int) [image/SourceBuffer.cpp:469]
#05: mozilla::image::RasterImage::OnImageDataAvailable(nsIRequest *,nsISupports *,nsIInputStream *,unsigned __int64,unsigned int) [image/RasterImage.cpp:1045]
#06: imgRequest::OnDataAvailable(nsIRequest *,nsISupports *,nsIInputStream *,unsigned __int64,unsigned int) [image/imgRequest.cpp:1191]
#07: ProxyListener::OnDataAvailable(nsIRequest *,nsISupports *,nsIInputStream *,unsigned __int64,unsigned int) [image/imgLoader.cpp:2789]
#08: mozilla::net::nsHttpChannel::OnDataAvailable(nsIRequest *,nsISupports *,nsIInputStream *,unsigned __int64,unsigned int) [netwerk/protocol/http/nsHttpChannel.cpp:7794]

Windows, Linux - Nightly 57.

This assertion is not very reproducible though it keeps appearing. I've been able to reproduce occasionally manually especially on Windows. I reproduced on Windows 10 32bit using https://www.charchinet.com/ this morning but was not able to reproduce again after the first time.

I am going to file this anyway in the hope that it will be clear as to what the issue is even if it is not easily reproducible.

Comment 1

[Bob Clary [:bc] (inactive)]

Attachment: urls

This includes most of the urls where this has been reproduced at least once. I've scrubbed some urls which may contain pii.

Comment 2

[Timothy Nikkel (:tnikkel)]

As far as I can tell the length argument (which we are asserting about) comes from the networking code.

Comment 3

[Milan Sreckovic [:milan] (needinfo for best results)]

Patrick, thoughts about comment 2?

Comment 4

[Patrick McManus [:mcmanus]]

you've got the cache on the stack so I'll redirect to michal..

but in general, this has been a common enough problem (especially with addons in the past!) that its worth handling it at runtime - so many things get their hooks into stream processing. see nshttpconnection.cpp:1685

Comment 5

[Michal Novotny [:michal]]

CacheFileInputStream::ReadSegments was called with aCount=0

Comment 6

[Michal Novotny [:michal]]

Other possibility is that we call the writer after aCount bytes were already written because we don't break the loop at http://searchfox.org/mozilla-central/rev/15ce5cb2db0c85abbabe39a962b0e697c9ef098f/netwerk/cache2/CacheFileInputStream.cpp#201.

Comment 7

[Michal Novotny [:michal]]

Attachment: fix

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/01c08db62e0f
Do not call aWriter with no data in CacheFileInputStream::ReadSegments, r=valentin

Comment 9

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/01c08db62e0f

Comment 10

[Ryan VanderMeulen [:RyanVM]]

Is there a user impact here that justified backport to Beta or can this ride the 58 train?

Comment 11

[Michal Novotny [:michal]]

AFAICS, calling SourceBuffer::Append() with aLength=0 should be no-op if it's not a first call. There should be no problem in release build.