Fix tabs not preserved in raw mode

[mohd-akram]

No description provided.

[gwsw]

Can you explain the motivation for this change? Normally less does not output tabs, but can be configured to do so via the -U option. In that case, currently less configures the tty driver to expand tabs to spaces. This PR configures the driver to output raw tabs instead. I think an argument could be made for either behavior. I can also see an argument to leave the TAB bits unchanged so that the tty driver does whatever the user has configured it to do outside less.

[mohd-akram]

The primary motivation is to be able to copy diffs when eg. using less as a pager for git (#152).

I think an argument could be made for either behavior.

IMO this behavior is more useful because the current behavior is a no-op for tabs in raw mode, and if one still wants expansion in raw mode, they can get it with --proc-tab. I also noticed the following comment, so it seems this was the intention:

less/screen.c

Line 373 in 26cfc73

* 4. \t is NOT expanded into spaces.

I can also see an argument to leave the TAB bits unchanged so that the tty driver does whatever the user has configured it to do outside less.

That's what my patch did initially, but then I saw:

less/screen.c

Line 631 in 26cfc73

s.sg_flags &= ~(ECHO|XTABS);

I guess that line would also need to be tweaked? I don't have a strong opinion on this part tbh, and since the user can control it, it doesn't hurt.

[gwsw]

I think the modified patch is ok. It was clearly wrong for the behavior to be different depending on whether TCGETA or TIOCGETP is supported.

However, I think the change should be made only to screen.c, not subsearch.lt nor term.c. The subsearch.lt file is not source code, it is just using an old copy of screen.c as a text file for one of the tests. And in term.c which is part of the lesstest program, the behavior should be deterministic, not dependent on how the user has set his terminal. The exact behavior (tabs expanded or not) doesn't really matter, but it does matter that the behavior is the same for every user, regardless of their tty configuration.

[mohd-akram]

That makes sense. Updated the patch to only change screen.c.