ftrace: Suspected parser errors cause majority of events after a certain point to not be listed

[felix-khlmnn]

Hello everybody,

Bug Description:

Some time ago I have documented a bug where the parsing of ascii ftrace files stop seemingly at random when going through the file. The result is that only the events up until this stop get displayed.

I have accidentally filed the same bug report for the main repo because I thought the ftrace component is included there and not here.

Further details can be found in #eclipse-tracecompass/org.eclipse.tracecompass#193.

The issue appears to be due to an event name starting with opening parentheses "(" which in turn seems to terminate the parsing process.

I'd like to work on a fix concerning this issue. Before I start my work on it, I'd like to know whether this is a known issue that you want to have fixed :)

Steps to Reproduce

1. Download the ZIP file below
2. Transform the binary file "long_program.dat" into a text file with trace-cmd report > long_program.txt
3. Load it into a Trace Compass project and compare the number of events of the binary file and the ascii file (Should be different due to error described above)
4. In the ascii file, search and replace all occurrences of "(udev-worker)" with "udev-worker"
5. Reload ascii file, compare again: Event number should be equal

Binary Trace File

Thanks in advance

Felix

[felix-khlmnn]

On further analysis, I'm sure that it has something to do with the Regex here, which decided whether a line is part of a valid ftrace or not: Regex line

[felix-khlmnn]

I have tested with regex101 and the issue seems to be in the way the "data" capture group is defined.

When we switch (?<data>[^\\)]*)(\\))? with (?<data>.*)$ regex101 gets a full match with the data in the correct fields.
I'll try to implement this soon and test it.

[nhths]

Any updates here?