Fix H5Fflush failure in hdf5_v1 printer with MultiNest#586
Open
Fix H5Fflush failure in hdf5_v1 printer with MultiNest#586
Conversation
The hdf5_v1 printer aborted with "H5Fflush returned error value (-1)... not a file or file object" whenever a scanner explicitly called flush() on an auxiliary print stream. MultiNest hits this on its very first dumper callback, where it flushes the "txt" (Posterior) and "live" (LastLive) RA streams. Root cause: only the primary HDF5Printer opens the HDF5 file and sets its file_id member; auxiliary printers inherit location_id, RA_location_id,and metadata_location_id from the primary but never initialise their own file_id. HDF5Printer::empty_sync_buffers() called H5Fflush(file_id, ...)unconditionally, so on an aux printer it was flushing an uninitialised hid_t. Fix: route the flush through primary_printer->file_id. primary_printer defaults to "this" for the primary and is set to the actual primary in the aux constructor, so the same call is correct in both cases. This also makes the dumper's flush() calls actually useful (they now push HDF5's internal buffers to disk) instead of erroring out. Closes #495.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The hdf5_v1 printer aborted with "H5Fflush returned error value (-1)... not a file or file object" whenever a scanner explicitly called flush() on an auxiliary print stream. MultiNest hits this on its very first dumper callback, where it flushes the "txt" (Posterior) and "live" (LastLive) RA streams. Simple reproduction of the issue: On the master branch, run a scan with
spartan.yamlset to use themultinestscanner and thehdf5_v1printer.Root cause: only the primary HDF5Printer opens the HDF5 file and sets its file_id member; auxiliary printers inherit location_id, RA_location_id,and metadata_location_id from the primary but never initialise their own file_id. HDF5Printer::empty_sync_buffers() called H5Fflush(file_id, ...)unconditionally, so on an aux printer it was flushing an uninitialised hid_t.
Fix: route the flush through primary_printer->file_id. primary_printer defaults to "this" for the primary and is set to the actual primary in the aux constructor, so the same call is correct in both cases. This also makes the dumper's flush() calls actually useful (they now push HDF5's internal buffers to disk) instead of erroring out.
A test with
spartan.yamlwithmultinestandhdf5_v1confirms that this fix works.Closes #495.