Re: Simplify target path generation?

From: David Abrahams (dave_at_[hidden])
Date: 2003-10-21 13:10:08

Rene Rivera <grafik666_at_[hidden]> writes:

> [2003-10-21] David Abrahams wrote:
>
>>"Kirill Lapshin" <klapshin_at_[hidden]> writes:
>>
>>> --- In jamboost_at_[hidden], Vladimir Prus <ghost_at_c...> wrote:
>>>
>>>> What do users think about it?
>>>> 
>>>> Two possible problems are:
>>>> 
>>>> 1. A bit longer paths. Not a real problem, I think.
>>>> 2. If one cpp is used by two exe, it will be recompiled twice. I
>>> think this 
>>>> will be quite rare.
>>>
>>> Fine with me -- paths already long enough to render them pretty much
>>> unusable :)
>>>
>>> So I have to rely on unit-test, symlink and stage anyway.
>>
>>Interesting viewpoint. Maybe if we can't make paths truly simple we
>>shouldn't worry about simplifying them even a little.
>>
>>IMO the only way to make them truly simple is to reduce paths like
>>gcc/debug/optimization-on to mangled names like "gdo".
>
> Or hash the directory name and only have something like:
>
> gcc/1234567/...

Sure, that could do the job, making the chance of a collision
infinitesimally small. Taking a cue from tinyurl.com we should
probably use a base-36 representation of the hash.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com