|
|
|
|
|
by ajross
1183 days ago
|
|
I'm not defending anything here. I'm saying that you guys launching verbiage like "unforgivable" or "casts serious doubt on" is unhelpful. Instead, try thinking about how problems like this can be avoided. Here's an example: the original change was a compatibility regression. Clearly there should have been a test of the original code somewhere that opened a file with "w" and validated that it was truncated per the documentation. And there wasn't. So one recommendation might be an audit of unit tests to verify that there's a process for getting from documented behavior to validated behavior. And importantly, there's no need to "doubt" or "forgive" to do that. |
|