D245 - The section written to explain D235 committed D235

status: decided

061-imports asks the resolver whether the platform has a symbol, and asking means loading the symbol's library. Ten libraries end the process when loaded (D236). The section opened them.

The first full run stops mid-walk, one library after libScePad; the next library in registry order is libSceVideoRecording. Twenty-eight sections' worth of measurement lost, to a library whose behaviour was already written down in this repository.

EXCLUDE could not have caught it. That mechanism matches check ids, and this is one check opening many libraries - so the thing that makes the announce-before-attempt property work for every other check does not apply here. Two changes make it apply:

• The library is named with a module record before it is opened, so a death names it.
• Libraries in data/hardware/crashers.txt are not opened at all. The Makefile derives the names from the record's check ids and passes them as OBS_UNSAFE_LIBRARIES.

A library that was not asked about emits no record. Writing unresolvable for it would say "the platform does not have this symbol" where the truth is "we did not ask", which is the precise conflation this whole section exists to undo. The count goes in the verdict, so the silence is accounted for rather than merely quiet.

Worth recording plainly: this was written immediately after reading D235, by someone who had just described the mistake, and it was made anyway. The rule "do not load a library as a side effect" is not hard to state and it is apparently easy to walk past when loading is the point of the check rather than an accident of it.