Clarify a comment regarding ffflus#115
Conversation
ffflus (on *-nix) eventually calls fflush. It does *not* call fsync or sync, and the data may not be on they physical device after this function returns. This change clarifies the comment to emphasize that.
|
Instead of changing a bunch of comments, maybe it should just call fsync since that seems to be the intention? |
|
Perhaps calling fsync is the right thing to do, but I'm certainly not familiar enough with the code yet to make such a change! At the moment, I just want comments that aren't misleading. Calling fsync could have substantial performance penalties for users of the library, and I don't think there is enough of a benchmark in place to be able to understand what the impact would be. |
|
The major question is: is the fflus useful as an intermediate layer that performs better on the memory-mapped FITS files of the mem: driver as long as fsync is not called? |
rmathar
left a comment
rmathar
left a comment
There was a problem hiding this comment.
Definitely a valid and useful update of the C-functions...
3 participants
ffflus (on *-nix) eventually calls fflush. It does not call fsync or sync, and the data may not be on they physical device after this function returns. This change clarifies the comment to emphasize that.