| Thread Previous • Date Previous • Date Next • Thread Next |
On 2010-08-13 07:37, Michael Hudson wrote:
The only thing I can think of is that the database isn't being reset when it should be, i.e. some test is managing to commit without the database being marked dirty. Does your branch cause tests to be run in a different order? I don't suppose it will be fun to try to answer that question :/
That's always a risk. It happens when a test runs a script and doesn't explicitly mark the database as dirty. Other database changes in the same test usually mask it.
Perhaps this would be an appropriate juncture to replace the tedious Popen rituals in tests with run_script calls(¹). Then make run_script mark the database as dirty and we've killed a whole class of silent-but-deadly test bugs.
Jeroen(¹) We do have several run_script functions, unfortunately, and that too would be a good thing to clean up.
| Thread Previous • Date Previous • Date Next • Thread Next |
