From: Andrey Semashev (andrey.semashev_at_[hidden])
Date: 2021-06-03 19:17:55
On 6/3/21 9:59 PM, Peter Dimov via Boost wrote:
> Andrey Semashev wrote:
>> On 6/3/21 9:30 PM, Peter Dimov via Boost wrote:
>>> Andrey Semashev wrote:
>>>> But that would mean one has to specify second order dependencies as
>>>> well for configure checks.
>> Ok, so it seems I'll have to add a check to enforce that the unified headers
>> directory exists. This means that we are still dependent on `b2 headers`, at
>> least until there is a CMake equivalent.
> "We" are not dependent on any such thing, and there will never be such a
> CMake equivalent. This directory will not be maintained in a CMake build.
Well, the only other alternative I see is globbing through all libraries
to collect their include paths. And that would have to be done in every
library, unless a common solution exists. I just looked at Boost.Atomic
and it also relies on the unified headers dir, and I'm pretty sure I'm
going to need it in Boost.Log. So the scanning will happen at least
three times per build, which is already not good.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk