Scrutinizer error tracking issue

[MrPetovan]

https://scrutinizer-ci.com/g/MrPetovan/friendica/issues/task/4176-scrutinizer

Log:

• January 16th 2018: 1691 issues
• January 19th 2018: 1681 issues
• February 7th 2018: 2018 issues (!?)
• February 13th 2018: 1964 issues (thanks @rabuzarus)
• February 14th 2018: 1898 issues
• October 22th 2018: 2222 issues (!!)
• December 25th 2018: 2459 issues (!!!) but the code quality score jumped from 2.9 to 3.54 out of 10!
• December 26th 2018: 2219 issues ([Scrutinizer] Replace deprecated method/function calls #6327)
• January 7th 2019: 2344 issues (because of the deprecation of Logger::log)

[MrPetovan]

@MrPetovan added Tedious Necessary labels a day ago

[annando]

In German you would say "Mühsam ernährt sich das Einhörnchen" (comparable to "Little by little, the bird builds its nest.")

[MrPetovan]

Here's how static analysis can concretely help a project code health in unexpected ways: When I updated dba::select() return type documentation to reflect the possibility of an array in case of limit => 1, the next Scrutinizer inspections showed a bunch of issues regarding type handling around the fact that we were calling array-only functions on the return data of dba::select(limit => 1).

It then hit me that we were doing two very different things with dba::select() and that it should probably be split into two distinct functions. This helped:

• Create dba::selectFirst() with a clear and simple purpose and non-ambiguous return types.
• Make dba::select() clearer and simpler with less conditions and non-ambiguous return types. I removed an input parameter ($params[only_query]) and a variable ($single_row) that didn't need to be moved to the new function.
• Simplify all the dba::select(limit => 1) calls to dba::selectFirst().

I probably could have come up with the change myself but the fact is that I didn't, and only my commitment to fix issues brought up by Scrutinizer finally prompted me to do it.

[MrPetovan]

I finally got to run Scrutinizer again after the TLS debacle surrounding git.friendi.ca, and unfortunately we jumped 300 issues since last February.

BUT the Scrutinizer Code Quality Score jumped from 2.91/10 to 3.47/10, so we must be doing something right.

[MrPetovan]

Currently the output of Scrutinizer isn’t useful, too many false positives, closing here.