BIND 10 #1964: regressions due to --enable-logger-checks
BIND 10 Development
do-not-reply at isc.org
Tue May 8 20:59:31 UTC 2012
#1964: regressions due to --enable-logger-checks
-------------------------------------+-------------------------------------
Reporter: jinmei | Owner:
Type: defect | Status: new
Priority: very high | Milestone:
Component: logging | Sprint-20120515
Sensitive: 0 | Keywords:
Sub-Project: Core | Defect Severity: N/A
Estimated Difficulty: 0 | Feature Depending on Ticket:
Total Hours: 0 | Add Hours to Ticket: 0
| Internal?: 0
-------------------------------------+-------------------------------------
This feature added in #1892 caused at least a couple of troubles.
The code also seems to have other potential issues.
First, it breaks compile when the check is disabled due to the unused
function:
http://git.bind10.isc.org/~tester/builder/BIND10-systest/20120508204501-MacOS/logs/build.out
Second, a related unit test fails at least on some environments
including mine:
{{{
[ RUN ] FormatterTest.mismatchedPlaceholders
log_formatter_unittest.cc:117: Failure
Death test: { isc::testutils::dontCreateCoreDumps();
Formatter(isc::log::INFO, s("Too many arguments in %1 %2"),
this).arg("only one"); }
Result: illegal return in test statement.
Error msg:
[ FAILED ] FormatterTest.mismatchedPlaceholders (4474 ms)
}}}
Third, the way dontCreateCoreDumps() is implemented is broken:
- we must not use an unnamed name space in a publicly shared header
file.
- liblog is a quite primitive library while testutils are generally
considered utilities for a more advanced modules such as data source
or the auth server. It's probably better to this set up in
lib/util/unittests
Combining these, I'd suggest declaring this function in a normal named
namespace in lib/util as a non inline function and defining it in a
separate .cc file. It will also work as a solution to the
unused-function problem.
Unfortunately the defect is quite big and it won't be a one-line fix.
We need a branch and normal review process at the highest priority.
Pushing this to the current sprint.
--
Ticket URL: <http://bind10.isc.org/ticket/1964>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development
More information about the bind10-tickets
mailing list