Fix AbstractInputStream#gets to not assume a_sep_string will appear…

[ListlessPanda]

The AbstractInputStream#gets has logic to check if the number of bytes within the @output_buffer are over the limit then sets over_limit. It seems that if a separator was passed in along with the number of bytes the code assumes the separator will appear within the bytes read.

This was picked up when streaming a CSV out of a ZIP where the CSV passes the a_sep_string as \n and the number of bytes as 8192 and the file contains a single column value that is longer than 8192 bytes

I made a change that if the separator wasn't found and we are above the limit it will return up to the number of bytes and increase the @pos by that amount.

(This is my first pull request please let me know if I need to do anything else)

[hainesr]

Hi @ListlessPanda,

Thanks for looking at this. I've been doing a load of updates to AbstractInputStream recently in #661. Please could you test your error case as you describe above against the code in #661? It may be fixed in there already and I plan to merge and release that soon.

[hainesr]

Looking at the tests I added in #661, I think I test the case you mention above, here:
https://github.com/hainesr/rubyzip/blob/af08a4754c6565fd92dff1d3356d5fe974aa6ee3/test/ioextras/abstract_input_stream_test.rb#L360-L371

You can see that some of the returned lines are truncated by the limit before a separator is found. Is this what you were fixing?

[hainesr]

I've now merged #661 and released as v3.3.0, so please can you test that and see if it fixes the issue you're seeing?

[hainesr]

I think this has been fixed in v3.3, so closing this PR, but please reopen if not the case and I will have another look.