On Mon, 12 Feb 2024 10:11:52 +0100 Pierre Gondois <pierre.gondois@xxxxxxx> wrote: > Ok right. I think I got confused due to the possibility to have multiple > output files when expanding 'trace-cmd split <TOP_COMMANDS> <INSTANCE_COMMANDS>' Yeah, I think we got on the tangent of having split become too powerful for one line. Having it only have one output name probably is the least confusing. It can be called multiple times if someone wants more complex parsing. > > I think the description should match this then: > trace-cmd split [OTHER_OPTIONS] [--top [-b name]] [-o output_file] [-B instance_name [-t]]* > [start-time [end-time]] > # The '-t' option can only be used once per command line > > or: > trace-cmd split <TOP_COMMANDS> <OUTPUT_COMMAND> <INSTANCE_COMMANDS> > > TOP_COMMANDS :: nil | --top TOP_PARAMS > TOP_PARAMS :: nil | -b name > > OUTPUT_COMMAND :: nil | -o output_file > > INSTANCE_COMMANDS :: nil | -B name INSTANCE_PARAMS INSTANCE_COMMANDS > INSTANCE_PARAMS :: nil | -t > # INSTANCE_PARAMS can only expand to '-t' once. > > I'm also not sure it is necessary to have a specific order for the > --top/-o/-B parameters. I should also work when they are unordered > if we limit the command to one output file: > - trace-cmd split -B foo -t --top -b old_top -o output.dat > - trace-cmd split -o output.dat -B foo -t --top -b old_top > - trace-cmd split -o output.dat --top -b old_top -B foo -t > should all be equivalent. Yeah, that's fine. Let's go with that. -- Steve
|