Re: [Cscope-devel] Vim issue and patch submission

[Ken S. <kg...@gm...>]

2011/2/20 Hans-Bernhard Bröker <HBB...@t-...>:
> On 17.02.2011 19:31, Ken Smith wrote:
>
>>>> I'm happy to learn that cscope already does what I want but so far, I
>>>> don't think -P does.  I'm happy to be proven wrong.
>>>
>>> You don't want to pass $(pwd) to -P.  It's meant to be used like this:
>>>
>>>        cscope -d -f /some/where/cscope.out -P/some/where
>>
>> This does work but I find it more straightforward to build the
>> database with absolute paths.  Is there any harm in that?
>
> There's always the off chance that people might be storing huge source tree,
> including their cscope databases, on network drives, which are mounted
> differently by different users.  -P survives that, absolute paths wouldn't.

Fair enough.  But I think the usual case is that every developer has a
working copy of everything they need checked out from version control
and generates a cscope database for it.  Also, back in the day, I
remember folks mounting network drives differently on different
computers but lately I don't encounter that.  Perhaps folks are coming
to the conclusion that a consistent environment is more valuable than
giving the flexibility and responsiblity of filesystem organization to
the end user.

I agree that the interface of cscope is not incorrect.  I assert that
it optimizes for the rare case and is redundant in the common case.

> Or one could build the database for one source tree, then use it to browse a
> modified copy of it.  That currently doesn't need a -P option at all.
>
>> But isn't it more convenient to simply build the database with
>> absolute paths and then you don't have to use -P to load each one of
>> them every time.  Indeed, the invocation is somewhat redundant don't
>> you think?
>
> Only if the paths for -f and -P were _always_ the same.  Which, as shown
> above, needn't be true.

I'm only talking about default behavior which I admit may be too late
since users probably already rely on the present default behavior.  I
am suggesting that -f path should assume -P `basename path`.  Let -P
override this default behavior.

In this case too, I think cscope's interface optimizes for the uncommon case.