GSB stream reader's `fill_value` doesn't do anything

[cczhu]

GSB stream reader's `fill_value` doesn't do anything

[mhvk]

@cczhu - but GSB can have missing data blocks, which would be noticed from the timestamp file. Still, since currently we do not do anything with it, I guess we might as well remove it for now.

[cczhu]

@mhvk that's true, but since payload-timestamp connections are inferred rather than embedded within the GSB file, is it even safe to keep reading after hitting a missing timestamp? It could be that the payload kept writing, or was partly written. I suppose we could do a complete file integrity check at the beginning to make sure the payload file sizes add up to the number of header lines, and then assume missing data only if everything does check out (which still leaves the possibility for some forms of time offsets to go unchecked)?

[mhvk]

"Needs clarification" since I think there is a decision to be made here: my current sense is that fill_value should be passed in at file opening time rather than for reading, since it is very unlikely one would want to change it half-way through (see #146). An advantage of that method is that fh.read() starts to become completely independent of what the underlying stream is. If we go that route, then for gsb the conclusion is simply to remove fill_value from its read method.

[mhvk]

fixed by #163 (by removing the fill_value for now).