Fixes#258
Scenarios:
* Launch Files
* Launch files with custom cwd
For each of the above scenarios:
* Test output
* Test breakpoints
* Test log points
* Test conditional breakpoints
* Test run to completion
(see #530)
os.kill() terminates the process immediately. However, we need the "disconnect" response to be sent (indirectly via the atexit handler). This PR switches to sys.exit() solve the problem. We still terminate the process early if disconnecting during a launch session.
(see #530)
We were sending the "disconnect" response only after closing the socket, meaning the response was never actually sent. This PR fixes that by ensuring the response is sent as late as possible, whether or not it is a "single session" situation.
(fixes#285)
We were not clearing things like breakpoints nor resuming threads when an "attach" session disconnnected. This caused problems, e.g. when disconnect happened with a thread stopped for a breakpoint. This PR fixes the problem by resetting all session-dependent debugger resources after the "disconnect" response has been sent (only for an "attach" session).
(fixes#448)
While #502 helped with the race between enable_attached() and adding breakpoints, there was still a small race left. This PR fixes it by CMD_RUN is not issued until after the breakpoints have been added.
* Test compat fixes for python 3.4 tests
* Enable python 3.4 CI in travis
* Fix for 3.4 missing text in traceback list
* Fix linter
* Add python 3.4 to classifiers
* Fix for line text
* Add flake8 configuration files
- .flake8.ci: addendum config for CI to make use of only.
* Add build badge for CI in VSTS
* Enable JUnit style test file output.
- convert_args return value simplified
- allow original style of unittest reporting as well
- remove unnecessary flags from junit xmlrunner
* Use argparse to collect runtime configuration of tests module
(see gh-448)
There is a race between attach and handshake completion that may result in breakpoints not being hit. This PR fixes that by waiting for handshake completion when waiting for attach.
* Fix issue where pydevd threads could end up being traced.
* Make sure that the stepping command is properly set when a thread is suspended. Fixes#501, #479
(fixes#512)
The XML library we are using treats a single-item list as a single item instead of a list. This breaks on_stackTrace(). This PR fixes the problem by converting xframes to a list before checking its size.
(fixes gh-243)
Under some corner cases the tests for pathutils fail, due to how Python's import system does not resolve casing on filenames. This PR fixes that.
(fixes gh-459)
Our exit handling was all over the place. I've refactored the code to properly separate concerns. One consequence is that this fixes wait-on-exit.
* Add set debugger property to support VS just my code
* Just my code initial code
* Limit continue to stepping and exceptions
* Improve skipping non-user code
* Fix tests
* Just my code shoudl skip ptvsd files
* fixes and updates (#1)
* Fix typo
* Use CMD_STEP_INTO_MY_CODE and CMD_SET_PROJECT_ROOTS
* Update tests to reflect the new pydevd exception breakpoint cmd args
* Fixes and improvements
* Simplify and refactor debug options
* Test fix and fix for debug property request handler for VS
* More fixes
* Test fixes
* Address comments
* Bunch of test fixes (#2)
* Test fixes
* Restoring auto-continue till pydev side work is done.
