Should indent-tabs-mode insert tabs only leading whitespace?

Should indent-tabs-mode insert tabs only leading whitespace?

[Daniel Colascione]

I recently debugged a weird interaction between make-mode and
ws-butler-mode that let to random tab characters being strewn in my
makefiles: ws-butler-mode uses move-to-column after save to restore
trailing whitespace on the line containing point. move-to-column calls
indent-to to actually do the move. And because makefile-mode sets
indent-tabs-mode, that indent-to inserts tabs sometimes, resulting in
stray tab files in places where they don't belong.

There are lots of fixes for this problem, but the fundamental issue seems
to be that indent-to inserts tab characters for column alignment *after*
non-whitespace characters on a given line. What if indent-to always
inserted spaces when there were non-whitespace characters between BOL and
point? That whitespace manipulation isn't really indentation.

Re: Should indent-tabs-mode insert tabs only leading whitespace?

[Stefan Monnier]

> There are lots of fixes for this problem, but the fundamental issue seems
> to be that indent-to inserts tab characters for column alignment *after*
> non-whitespace characters on a given line. What if indent-to always
> inserted spaces when there were non-whitespace characters between BOL and
> point? That whitespace manipulation isn't really indentation.

AFAIK `indent-to` is a low-level function that is used for indentation
indeed but is also used for alignment and many other purposes, so
changing it as you suggested will likely lead to undesired behaviors in
some cases.

So I think this should be fixed elsewhere or at least be conditional on
some new variable setting (e.g. `indent-tab-mode` set to `only-at-bol`).


        Stefan