I think ignore_user_abort(true) should be used inside ParallelSave::buildFullFileFromChunks method merges the chunk files. If the client disconnects during that process some of the chunks gets deleted and a partialy merged file is created. By ignoring user abort the process could finish without any problem.
This also applies to ChunkSave. During the merge of the last chunk user abort could be ignored. With this, the last merge and the move could finish without interruption.
This could lead to some inconsistency after the client reconnects.
Maybe this could be configurable and being disabled by default.
I think
ignore_user_abort(true)
should be used insideParallelSave::buildFullFileFromChunks
method merges the chunk files. If the client disconnects during that process some of the chunks gets deleted and a partialy merged file is created. By ignoring user abort the process could finish without any problem.This also applies to
ChunkSave
. During the merge of the last chunk user abort could be ignored. With this, the last merge and the move could finish without interruption.This could lead to some inconsistency after the client reconnects.
Maybe this could be configurable and being disabled by default.