refactor callinfo to simplify ctor magic #4299
Conversation
nicoddemus
changed the title
|
CI failures appear to be unrelated. |
|
Will squash/rebase later today @nicoddemus PS this is roughly what i want to enable for nodes AS well |
|
How is #4292 meant to look with this? diff --git i/src/_pytest/runner.py w/src/_pytest/runner.py
index 0d6e3dc3..045eb72d 100644
--- i/src/_pytest/runner.py
+++ w/src/_pytest/runner.py
@@ -15,6 +15,7 @@
from .reports import CollectReport
from .reports import TestReport
from _pytest._code.code import ExceptionInfo
+from _pytest.outcomes import Exit
from _pytest.outcomes import skip
from _pytest.outcomes import Skipped
from _pytest.outcomes import TEST_OUTCOME
@@ -190,10 +191,13 @@ def check_interactive_exception(call, report):
def call_runtest_hook(item, when, **kwds):
hookname = "pytest_runtest_" + when
ihook = getattr(item.ihook, hookname)
+ reraise = (Exit,)
+ if not item.config.getvalue("usepdb"):
+ reraise += (KeyboardInterrupt,)
return CallInfo.from_call(
lambda: ihook(item=item, **kwds),
when=when,
- reraise=KeyboardInterrupt if not item.config.getvalue("usepdb") else (),
+ reraise=reraise,
)? Or would I think this tights Exit and KeyboardInterrupt just a bit more (as it currently stands)?! |
|
@blueyed correct - the signaling should be layered differently |
|
How then? (please be more verbose) |
|
@blueyed my own investigation on that is not finished as the layering and the structure of calls, hooks and nodes is intertwined with this detail issue and its not clear how to layer and integrate them - i took a while to think if something comes to mind, but right no good solution seems to come to it |
|
But you would like to keep this PR as is? (but IMHO given the above diff it does not really make it any better than https://github.com/pytest-dev/pytest/pull/4292/files#diff-4d6d922c3b97a9eae157f8f653627b13 would have done it) |
|
thanks for the node, i got sidetracked ^^ |
RonnyPfannschmidt
force-pushed
the
callinfo-sanity
branch
from
c434f67
to
5e6553c
Compare
|
pushed my squash/rebase - i might have invalidated the repr while in call test and should take a look at removing that again - will do more after i fetched the kid from daycare |
Codecov Report
@@ Coverage Diff @@
## features #4299 +/- ##
============================================
+ Coverage 95.81% 95.83% +0.02%
============================================
Files 111 111
Lines 25034 25038 +4
Branches 2462 2463 +1
============================================
+ Hits 23987 23996 +9
+ Misses 737 735 -2
+ Partials 310 307 -3
Continue to review full report at Codecov.
|
RonnyPfannschmidt
force-pushed
the
callinfo-sanity
branch
from
5e6553c
to
1b07320
Compare
|
yup, had to take it out again |
RonnyPfannschmidt
changed the title
testing/test_runner.py
Outdated
| assert ci.when == "123" | ||
| assert ci.result == 0 | ||
| assert "result" in repr(ci) | ||
| assert repr(ci) == "<CallInfo when='123' result: 0>" | ||
| ci = runner.CallInfo.from_call(lambda: 0 / 0, "123") |
There was a problem hiding this comment.
This test should not be changed.
RonnyPfannschmidt
changed the title
RonnyPfannschmidt
force-pushed
the
callinfo-sanity
branch
from
1b07320
to
13380e3
Compare
|
appveyor is codecov related |
RonnyPfannschmidt
changed the title
|
@RonnyPfannschmidt this is conflicting with the recent merge for |
RonnyPfannschmidt
changed the title
|
will do so either in the late afternoon or tommorow |
RonnyPfannschmidt
force-pushed
the
callinfo-sanity
branch
from
13380e3
to
847eace
Compare
RonnyPfannschmidt
changed the title
No description provided.