1. Mixed mode with current code is really hard to implement, so I defer this to separate repository. 2. Yes. With pure profile mode we haven’t any connection to debugger. The only way is manual importing cache grind file (already done). To simplify this I would like to introduce dir scanner (especially for SSH connections), where user can easy import via double click. Se also how work webgrind. 3.In this way we working right now. xdebug can populate aggregated files, but I didn’t test this yet. 4. We haven’t response header, especially in pure profile. We only have list of cache grind files and can read it’s creation date. If user select correct name schema, we can detect url, date etc..[1] 5. It’s related to point 2. We should watch output dir via SSH session, and download file on demand.
Current patch state: 1. Support cache grind manual import with state saving 2. Support php executable profiling session. PHPUnit support should be done via separate thread, mostly due nightmare in launchdelegates
What we read from file: 1. Complete call tree with timestamps 2. List of files / classes / functions and it’s positions
UI allow: 1. Navigate over call tree 2. Navigate to call file:position 3. Navigate to definition file:position 4. Display invocation statics (call and callers list)
I would like to drop “[WIP]” and merge patch as is and continue work via separate bugs like: 1. Support xdebug 2.6 memory statistics 2. Reduce UI freezing as much as possible - Execution Statistics tree it’s extremely slow 3. Improvements around Invocation Statistic tables. I thinking about docking inside Execution Statistics, and show call/callers based on current selection. 4. Model improvements to allow flag call tree positions as “internal function call”, “file includes include” etc.. and filters for that 5. output_dir scanner 6. Support xdebug trace file 7. General performance improvements, currently pdt store everything in memory. Disk store is must have 8. Dark theme improvements 9. probably more … this patch opens another window in my head :P
Hi David,
1. Yes, I'd prefer a checkbox in "profile" launch instead of
adding a new "profile and debug" profile
2. Would that work also when the server is not directly
controlled by PDT (like Apache or NGINX)?
3. I'd prefer a profile session for each request: profiling is
very helpful to see where a request is taking a lot of time, and
mixing requests in the same session would make it
harder.
4. What about adding a filter for the Content-Type response
header? We could also add a filter for paths (eg "I want to profile
the requests that starts with "/path/to/uri/resource/")
5. I don't understand ;)
--
Michele
_______________________________________________
pdt-dev mailing list
pdt-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/pdt-dev
|