> > AIR users will still have to use the DebuggableWorker class as base 
> > class for their workers and the main app as well though. 
> I'm not sure we can convince all developers to rewrite their workers, 
> but I hope no fatal error happens if fdb fails to suspend worker.

FDB not being able to halt the player, won't IntelliJ hang up in that case ?

> - I have to suspend at least one worker to be able to send 'break' 
> command. Fdb suspends/sets breakpoint/resumes other workers if needed, 
> at the end all workers are in the same state as before break command
> - the same about 'delete' command.

Correct.
Note: For the "delete" command, you can now delete unresolved BP, that wasn't 
the case before.

> - when breakpoint is hit, only one worker is paused.

Not correct, look at the debugTickler() at DebuggableWorker.as:56, if I set a 
BP there and somewhere else in another worker and this last one is hited first, 
it doesn't mean that the first one won't be hited in between, 2 workers can 
then be in suspend mode because I received 2 break events even though FDB will 
process them one at time.

> I have one more request for the IDE integration. Using part of file name 
> in 'break' command is dangerous, more than one file may match. Even 
> using full file name is not strict enough: we have a lot of users who 
> had the same-named files in different packages. Currently IntelliJ IDEA 
> tries to use file id instead of file name whenever possible, but with 
> your fixes I start thinking about a different approach. Will it be easy 
> to support relative path in 'break' command syntax? In this case I'll be 
> sending path relative to src root, e.g. 'break 
> flash/display/Sprite.as:99'? Implementation should normalize slashes in 
> known file paths as I'd like to always pass '/' whereas paths returned 
> by Flash Player may contain '\'.

Ok, implemented, will commit today, you can now do:

break mx\core\DebuggableWorker.as:56
or
break mx/core/DebuggableWorker.as:56

> > I just wonder, is it easily doable to modify the Breakpoint UI like 
> > what is done for Java and adding a combobox to access threads on top 
> > of the Frame list ?
> Shouldn't be extremely hard.

Can't wait to see that :-)

Thanks,
Frédéric THOMAS

> Date: Wed, 28 May 2014 22:19:10 +0400
> From: alexander.doros...@jetbrains.com
> To: dev@flex.apache.org
> Subject: Re: [FDB] Integration
> 
> On 28.05.2014 16:06, Frédéric THOMAS wrote:
> > AIR users will still have to use the DebuggableWorker class as base 
> > class for their workers and the main app as well though. 
> I'm not sure we can convince all developers to rewrite their workers, 
> but I hope no fatal error happens if fdb fails to suspend worker.
> > I had no problems using it with IntelliJ, waiting for more tests / 
> > feedback from you though :-) 
> It really seems to work! Please check if I get all correctly:
> - I have to suspend at least one worker to be able to send 'break' 
> command. Fdb suspends/sets breakpoint/resumes other workers if needed, 
> at the end all workers are in the same state as before break command
> - the same about 'delete' command.
> - when breakpoint is hit, only one worker is paused.
> 
> I have one more request for the IDE integration. Using part of file name 
> in 'break' command is dangerous, more than one file may match. Even 
> using full file name is not strict enough: we have a lot of users who 
> had the same-named files in different packages. Currently IntelliJ IDEA 
> tries to use file id instead of file name whenever possible, but with 
> your fixes I start thinking about a different approach. Will it be easy 
> to support relative path in 'break' command syntax? In this case I'll be 
> sending path relative to src root, e.g. 'break 
> flash/display/Sprite.as:99'? Implementation should normalize slashes in 
> known file paths as I'd like to always pass '/' whereas paths returned 
> by Flash Player may contain '\'.
> > I just wonder, is it easily doable to modify the Breakpoint UI like 
> > what is done for Java and adding a combobox to access threads on top 
> > of the Frame list ?
> Shouldn't be extremely hard.
> 
> Alex
                                          

Reply via email to