Don't insist on reading a full buffer from the input file.

Don't loop trying to read a full buffer from the input file.

If you're reading from a file, on UN*X or Windows, you should get the
entire read count unless you're fewer than buffer-size bytes from the
end of the file, in which case you should get what remains in the file.

If you're reading from a pipe, however, that could cause you to block
longer than necessary waiting for a full buffer rather than just for the
next chunk of data from the pipe - which might not be a bufferful, if
the program writing to the file is itself writing less-than-bufferful
chunks, as may be the case in, for example, a pipeline coming from a
live capture and with the intent that TShark display the packets as they
arrive.

While we're at it, if we're trying to do a seek and the seek takes place
within the buffer of uncompressed data, just adjust the position within
that buffer for forward seeks as well as backward seeks; this
substantially reduces the number of ws_lseek64() calls when making a
sequential pass through the file in Wireshark (e.g., running a tap or
filtering the display) and, as we purge the buffer after the
ws_lseek64(), substantically reduces the number of ws_read() calls in
that situation as well.

Have a data structure for a file data buffer, and use it for both the
"input" (compressed data) and "output" (uncompressed data) buffers.
Rename raw_read() to buf_read(), as it reads into a buffer.

Change-Id: I7982b3499a7613a993913a6db887054730764160
Ping-Bug: 14345
Reviewed-on: https://code.wireshark.org/review/25358
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>

0 files changed, 0 insertions, 0 deletions