Re: [jamboost] EXIT and errors.error

From: Rene Rivera (grafik666_at_[hidden])
Date: 2002-11-12 12:02:56

[2002-11-12] David Abrahams wrote:

>Vladimir Prus <ghost_at_[hidden]> writes:
>
>> David Abrahams wrote:
>>> Vladimir Prus <ghost_at_[hidden]> writes:
>>> 
>>
>>>>Backtrace, on the other hand, makes sense only
>>>>for developers, but not for end users.
>>> 
>>> 
>>> I'm inclined to err on the side of verbose error output until users
>>> tell us they can't handle it. Boost.Build v1 went the other direction,
>>> and users suffered.
>>
>> Convinced. Then, we need, for the sake of testing, introduce an option
which
>> makes errors.error do the same as print.wrapped-text.
>
>I disagree that it needs to do any message rewrapping.

I agree with you on that, for one simple reason... When running bjam in an
IDE/editor that parses the output it makes it possible to do error selection
so that the offending source/line shows in the editor.

> In fact, I have
>always had reservations about print.wrapped-text, even for
>documentation. People generally write sensible-looking comments, and
>sometimes the line breaks they choose are significant. There's no
>reason to think things will look better if they're re-wrapped. 

On this I disagree... I think documentation output, to console, should be
re-wrapped, just as it would be if it was HTML instead. I know people
generally write sensible-looking comments, but that is subjective. They
write sensible comments on their editor, which invariably is always set to
different proportions to everyone else. So people who use variable width
fonts will write longer lines than fixed width editing people causing
strange and inconsistent output to whoever reads the documentation. And
documentation should be the one thing that is the most consistent in a
program, as it's usually the first interface to your program people see.

-- grafik - Don't Assume Anything
-- rrivera_at_[hidden] - grafik_at_[hidden]
-- 102708583_at_icq - Grafik666_at_AIM - Grafik_at_[hidden]