Waiting for any non-zero duration currently waits for at least one timer
cycle (10 ms), which is especially expensive during early boot.
The current workaround of simply reading the status 14 times seems really
suspicious although the osdev wiki documents it, but let's see how well it
works on real hardware, it's probably good enough.
Try to determine the initial selected drive to save one drive selection.
Pagination code deals with two indices for video modes, one relative to
the start of video modes array and one relative to the start of current
page. Previously when displaying the list of modes, the video mode array
would be accessed using the one relative to the start of the current
page, meaning that pages 2 and onwards displayed repeats of the video
modes on page 1. This changes the the display code to use indices
relative to the start of the video modes array when accessing the array.
Previously setting a mode from the command line would only work if it
was one of the pre-set resolutions offered by the driver. If the driver
supported custom resolutions and a user tried to set one on the command
line, chvideomode(1) would instead launch in interactive mode.
Previously chvideomode(1) used "WIDTHxHEIGHTxBPP" on the command line
while interactively it used "BPP x WIDTH x HEIGHT". Change everything
to "WIDTHxHEIGHTxBPP".
Previously if argv ended with a long option that required an argument
followed by said argument, getopt_long(3) would generate an error saying
that the option requires an argument even though it was provided. This
was because the comparison of optind against argc did not account for
the fact that optind had already been incremented, causing an
off-by-one.
tix-port(8) parses the diff(1) output, to remove the "Only in …"
informational messages. As they are meant to be human-readable, not
machine-readable, they are translated to different languages, which
causes tix-port(8) to miss them. This patch sets LC_ALL=C for the diff
invocation, so that the messages are always in English.
