Re: [boost] Boost Digest, Vol 6513, Issue 3
Peter Dimov writes:
Marshall Clow wrote:
I must concur with Ville here; I have had several people tell me they won?t use Boost because ?it changes with every release?.
Boost does change with every release, and I still keep using 1.38 in several projects of mine, but these changes are not limited to ABI at all. Everything changes, you have to retest the whole lot, for correctness as well as performance.
And in general, expecting ABI stability from a collection of predominantly header-only libraries is setting one up for severe disappointment.
The standardized libraries are more header-only than Boost so am not sure how this isn't a rap on them also. Somehow C++ has become successful in spite of this.
Brian Ebenezer Enterprises - In G-d we trust. https://github.com/Ebenezer-group/onwards
On 3/12/2020 10:03 am, Brian Wood wrote:
Peter Dimov writes:
And in general, expecting ABI stability from a collection of predominantly header-only libraries is setting one up for severe disappointment.
The standardized libraries are more header-only than Boost so am not sure how this isn't a rap on them also. Somehow C++ has become successful in spite of this.
C++ prioritises performance, at the price of handing you a great many footguns with which to shoot yourself if not handled carefully. A great many developers are willing to make that particular tradeoff.
-----Original Message----- From: Boost
On Behalf Of Gavin Lambert via Boost Sent: 6 December 2020 21:56 To: boost@lists.boost.org Cc: Gavin Lambert Subject: Re: [boost] Boost Digest, Vol 6513, Issue 3 On 3/12/2020 10:03 am, Brian Wood wrote:
Peter Dimov writes:
And in general, expecting ABI stability from a collection of predominantly header-only libraries is setting one up for severe disappointment.
The standardized libraries are more header-only than Boost so am not sure how this isn't a rap on them also. Somehow C++ has become successful in spite of this.
C++ prioritises performance, at the price of handing you a great many footguns with which to shoot yourself if not handled carefully.
A great many developers are willing to make that particular tradeoff.
Developers should also factor-in and plan and budget ahead for a completely separate process of rebuilding and retesting and redocumenting each time a new revision is felt necessary. This might not be every Boost release, nor every compiler release. This will cost manpower, hardware and management, but the ultimate benefit is better quality and better performance. I find it extraordinary that anyone with an aspiration for quality is content with using obsolete bug-ridden compilers a decade old. Is this one reason why so much software 'sucks'? Paul
participants (3)
-
Brian Wood
-
Gavin Lambert
-
pbristow@hetp.u-net.com