Analysis TODO

[hainest]

• TODO why does number changed go up?

  • Just variance. No real reason.

• TODO TIM cross correlate 41 two predictor cases against the _chk cases found. When symbols says not compatible and libabigail says they are - what percentage are these _chk cases?

  • I'm working on this now

• TODO TIM THIS MANY Libabigail fails to report any reason for the incompatibility (e.g., the return value is non-zero without terminal output). TIM WHEREVER THIS GOES MAKE SURE TO ADD THAT WE FILTER LIBABIGAIL CASES - something something percentages don’t add up?

  • This has been accounted for by including all cases where the predictors disagree

• TODO TIM unique count of these to look at disjoint sets.

  • Results for this were added in Scripts for making paper numbers #11

• TODO some report on function changes/adds/removes

  • Results for this were added in Scripts for making paper numbers #11

• abi laboratory cases that fail with message "ERROR: can't find debug info in object" should be excluded from analysis. (completed in Exclude libraries where ABI lab couldn't find debuginfo files #14)

---

Items to add to report:

• Results from 2,3,4,5

• Output from 'agreement.py'

libabigail incompatibilities for changes in

• vtable
• enumerator
• ?? (Ben said he'd get me a list)