Fixes performance tests reporting completion times that are a little too early (1 second in some warm cache tests).
Could also fix any bugs in user applications where they use ::computeLoadProgress to determine loading completion.
Ex. waiting for loading before taking a screenshot, some parts of the screenshot might not have loaded to final detail
This fixes a bug in the mentioned function, where this->_workerThreadLoadQueue.size() is being used to calculate tile completion, but it's always 0, because it is cleared at the end of every ::update_view. The fix is to use the saved value in _updateResult.
Discovered while working on PR #779.
Also relates to PR #733, which has a similar fix. I wouldn't be surprised if this was the original cause, but that PR just happened to fix it too. Regardless, the older PR is still valuable. Counting "tiles kicked" as current progress sounds reasonable.
Fixes performance tests reporting completion times that are a little too early (1 second in some warm cache tests).
Could also fix any bugs in user applications where they use
::computeLoadProgress
to determine loading completion. Ex. waiting for loading before taking a screenshot, some parts of the screenshot might not have loaded to final detailThis fixes a bug in the mentioned function, where
this->_workerThreadLoadQueue.size()
is being used to calculate tile completion, but it's always 0, because it is cleared at the end of every::update_view
. The fix is to use the saved value in_updateResult
.Discovered while working on PR #779.
Also relates to PR #733, which has a similar fix. I wouldn't be surprised if this was the original cause, but that PR just happened to fix it too. Regardless, the older PR is still valuable. Counting "tiles kicked" as current progress sounds reasonable.