poboiko added a comment.
In D23008#508782 <https://phabricator.kde.org/D23008#508782>, @bruns wrote: > @poboiko - is the problem you describe purely theoretical, or did you actually see it happen? > > Multiple triggers of `slotNewInput` are **not** a problem, the code flow reaches the `m_tr = new Transaction(...)` only iff a batch has been read. And batches are only issued when the previous one has been signaled completed with the 'B' batch end marker. To be completely honest, it //is// theoretical. I do see it should not happen. Provided everything works as expected. There were just too many examples (even inside baloo), when critical things didn't work as expected, and got silently ignored due to asserts. I thought it won't hurt getting rid of those, replacing them with proper checks. Even if it's checks of stuff that's impossible. I did see extractor hanging though, without reliable way to reproduce it, and didn't have time to debug it when it happened. That's why I've started lurking around extractors code in the first place, looking for stuff like that. REPOSITORY R293 Baloo REVISION DETAIL https://phabricator.kde.org/D23008 To: poboiko, #baloo, bruns, ngraham Cc: anthonyfieroni, broulik, kde-frameworks-devel, LeGast00n, fbampaloukas, domson, ashaposhnikov, michaelh, astippich, spoorun, ngraham, bruns, abrahams
