Subject: Re: [boost] CMake Announcement from Boost Steering Committee
From: Robert Ramey (ramey_at_[hidden])
Date: 2017-07-19 15:54:45
On 7/18/17 11:57 PM, Gavin Lambert via Boost wrote:
> On 19/07/2017 14:41, Robert Ramey wrote:
>> One thing that always seems to get lost in this discussion is the
>> distinction between library developers and library users. Library
>> users need a simple way to import libraries. If one where to make (or
>> if it already exists) most library users would be satisfied if all
>> that had to do was something like FindBoostSerializaton or something
>> like that. Since they are downloading binaries, they wouldn't care how
>> boost was built etc.
> That doesn't necessarily follow.
> From that perspective, there's two kinds of Boost library users
> (ignoring those who want to actually modify Boost code):
> 1. The kind that just downloads prebuilt binaries and wants to use them.
> This group could be amply served by providing a pkgconfig file for
> Linux and a props file for Windows+VS -- though since ultimately in both
> cases it's mostly just specifying where the files were put (which could
> be different for each user) this is probably more of a documentation
> thing (providing a template where they fill in the actual path
> themselves) than anything else. I suspect the Linux distribution
> package maintainers probably already do this sort of thing; if you want
> to use system-provided Boost then it Just Worksâ¢.
> 2. The kind that wants to build the libraries themselves.
> There's many reasons for this; some want to use them on compilers or
> environments for which prebuilt binaries are not available; some want to
> tweak the settings for some particular library; some just want
> confidence that they have the actual source for the binary. These
> people mostly just want a simple command to build everything (or
> whatever subset they require). For the most part, they don't care how
> it happens or what tool is used, but the fewer external dependencies the
> better (as fewer things to go wrong).
> I've made some assumptions, of course, but I believe this is true.
FWIW - in 15 years, getting feedback from users - mostly bugs or
confusions about the library, I have never once found anyone who has
actually run the tests for the serialization library on their own
system. The only testing, other than my own, that I have been able to
discern has been on the boost testing matrix.
Also note that I have included CMake files for building and testing the
serialization library for several years and I have never received any
feedback on this. This leads me to conclude one of the following
a) the CMake version of the build test is absolutly correct and the
library tests always pass so there is nothing ever to report.
b) No one has ever used these CMake files.
> Unsubscribe & other changes: