Segfault when running flvdump on file

[GoogleCodeExporter]

What steps will reproduce the problem?
1. flvdump Baseball+Tennis.flv

What is the expected output? What do you see instead?
I expect the Metadata contents, and the tag stats. Instead, flvdump
segfaults after displaying metadata contents.

What version of the product are you using? On what operating system?
1.0.5, hardy i386

Please provide any additional information below.

Magic: FLV
Version: 1
Has audio: yes
Has video: yes
Offset: 9
--- Tag #1 at 0xD (13) ---
Tag type: Meta
Body length: 378
Timestamp: 0
* Metadata event name: onMetaData
* Metadata contents: {
    duration => 29.963000
    videodatarate => 329.760037
    lastkeyframetimestamp => 27.628000
    lastkeyframelocation => 1226111.000000
    creator => YouTube, Inc.
    metadatacreator => YouTube Metadata Injector.
    haskeyframes => true
    hasmetadata => true
    keyframes => {
        times: [
            0.000000
            1.034000
            9.376000
            17.718000
            26.059000
            27.628000
        ]
        filepositions: [
            448.000000
            5832.000000
            357005.000000
            768728.000000
            1154872.000000
            1226111.000000
        ]
Segmentation fault

Original issue reported on code.google.com by tile...@gmail.com on 27 May 2008 at 2:51

Attachments:

• Baseball+Tennis.flv

[GoogleCodeExporter]

I could reproduce the segfault on hardy heron, I'm investigating into this.

The Windows version does not crash, and reports some weird values for the 
metadata.
It seems like the file is kind of broken, even though it should not make 
flvdump crash :)

I've seen a few files from Youtube with strange values. It seems their metadata
injector isn't fully compliant with the AMF0 spec.

Original comment by marc.noi...@gmail.com on 27 May 2008 at 5:53

[GoogleCodeExporter]

Original comment by marc.noi...@gmail.com on 27 May 2008 at 5:54

• Changed state: Started

[GoogleCodeExporter]

The crash has been fixed, and flvmeta 1.0.6 will soon be released as a result.

The main cause of the bug was that fprintf under Linux didn't check for NULL 
pointers
passed as arguments.
Therefore, I now use the string size to limit the numbers of characters to be 
printed.

Original comment by marc.noi...@gmail.com on 28 May 2008 at 10:13

• Changed state: Verified