Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 313010

Summary:

Only breakpoints set during a debug session do actually break

Product:

[WebTools] JSDT

Reporter:

Michael Schneider <michschn>

Component:

Debug

Assignee:

Simon Kaegi <simon_kaegi>

Status:

RESOLVED FIXED

QA Contact:

Simon Kaegi <simon_kaegi>

Severity:

normal

Priority:

CC:

david_williams, Michael_Rennie, thatnitind

Version:

3.2

Flags:

david_williams: pmc_approved+
Michael_Rennie: review+
thatnitind: review+

Target Milestone:

3.2 RC2

Hardware:

OS:

Windows 7

Whiteboard:

PMC_approved

Attachments:

Description	Flags
proposed patch	none

Description Michael Schneider

2010-05-15 07:32:20 EDT

a video tells more than a thousand words :)
http://www.poshnet.ch/jing/2010-05-15_1325.swf

The breakpoints are only active when I set them during a debug session. However, as shown in the video, I'm positive that the filename I provide to rhino when compiling scripts can actually be mapped back to a workspace file. I use filenames like:
{"Bundle-SymbolicName":"com.ibm.team.apt.web.ui.tests", "path":"\\resources\\gridviewer", "name":"GridViewerUpdateTest.js"}
The line number I specify is 1.

This happens on the latest version from CVS

Comment 1 Simon Kaegi

2010-05-17 11:28:36 EDT

I was able to reproduce this fairly easily as part of today's test pass. We need to investigate a fix this for RC2.

Comment 2 Simon Kaegi

2010-05-19 14:38:06 EDT

Created attachment 169179 [details]
proposed patch

Comment 3 Michael Rennie

2010-05-19 16:16:10 EDT

+1 perfect.

Comment 4 Simon Kaegi

2010-05-19 17:21:17 EDT

As it stands the mapping between workspace resources and script uris is broken and this prevent breakpoints from working in mainline cases. As a result any breakpoints set in the workspace before a script is loaded by the vm being debugger are being ignored.

The breakage occured because the code path where we were requesting the breakpoint set were different based on whether the script was already loaded before the breakpoint was set or vice-versa.

This patch corrects workspace relative path usage so that all script paths are valid URIs that either map perfectly to a workspace resource or else are treated as external source. The same codepath is now used for both breakpoints set before and after the debugger loads a script.

Comment 5 David Williams

2010-05-19 20:20:55 EDT

pretty large change, but I agree, mainline case that deserves to work right. 

Thanks.

Comment 6 Simon Kaegi

2010-05-19 21:56:21 EDT

Fixed in HEAD

Thanks all.

Comment 7 Simon Kaegi

2010-05-19 21:56:42 EDT

Comment 8 Michael Schneider

2010-05-20 11:37:55 EDT

this seems to work now, however the JavaScript load breakpoints are acting weired now. I do not know if this is an issue with my setup or not. Do script load breakpoints still work for you?

Comment 9 Simon Kaegi

2010-05-20 13:33:06 EDT

(In reply to comment #8)
> Do script load breakpoints still work for you?

Thanks and sorry I missed this in my tests. Absolute URIs do work, but the more typical workspace one's do not as I missed this one. I've opened bug 313782 for RC3 and attached a patch and this now works for me.