For the 101 minutes video: https://www.youtube.com/watch?v=1OEu9C51K2A
For the C++ Core Gudilelines: https://github.com/isocpp/CppCoreGuidelines
For the Microsoft GSL: https://github.com/isocpp/CppCoreGuidelines
Bjarne Stroustrup talks basically on the Guidelines Support Library (GSL). I took the following notes while watching the excellent talk:
02:55 Very few people write modern C++ code.
04:40 Inventing a new language in order get a better and simpler
one is a bad idea!
08:50 We don't like coding rules. They are aimed to prevent
accidents: basically "don't do" lists. Not aimed for quality
code.
09:50 Coding rules recommends restrictions on C++ features.
10:00 There are a lot of bad advice!
12:07 I don't understand the MISRA C++ pointer rules: they are
miserable!
17:56 Language subsetting doesn't work.
18:15 Write high-level code via libraries using all the language
features.
19:10 After you build the superset, i.e. language + libraries, you
subset.
19:30 What we want is "C++ on steroids".
23:40 Tool support, i.e. static analysis, is essential.
25:10 No resource leaks: RAII, i.e. no naked new and delete!
32:50 GSL: owner is for fixing the dangling pointer nightmare.
33:15 Owner is intended to simplify static analysis.
37:00 Besides dangling pointers, there are other pointer misuses:
range errors: fix it with array_view.
nullptr dereferencing: fix it with non_null.
37:18 We want to minimize run-time checks, because we want to run
fast!
37:50 My least favorite interface:
void f(int *p, int n);
why?
So common!
So error-prone!
Fix it with array_view.
40:15 nullptr problems: fix it with not_null.
41:30 not_null is for minimising run-time checking!
42:20 "Smart pointers" are overused to overcome ownership and
dangling pointer problems.
43:12 As the dangling pointers are eliminated by tools, you don't
need to use them!
47:30 Too many rules: the tools should know the rules and will
point you the relevant ones.
49:15 There are too many rules and it doesn't work me! I want lots
of rules.
50:00 I hate camel case.
51:20 We aren't unambitious:
Aim type and resource safety.
Consider all 4+ millions C++ programmers.
Zero-overhead principle.
51:50 We aim to change the way we write code.
53:52 Help wanted! There are currently only two editors: me and
Herb Sutter.
55:02 The basic C++ model is now complete.
57:20 Starts to answer the questions.
For the C++ Core Gudilelines: https://github.com/isocpp/CppCoreGuidelines
For the Microsoft GSL: https://github.com/isocpp/CppCoreGuidelines
Bjarne Stroustrup talks basically on the Guidelines Support Library (GSL). I took the following notes while watching the excellent talk:
02:55 Very few people write modern C++ code.
 |
| 03:36 |
one is a bad idea!
08:50 We don't like coding rules. They are aimed to prevent
accidents: basically "don't do" lists. Not aimed for quality
code.
09:50 Coding rules recommends restrictions on C++ features.
10:00 There are a lot of bad advice!
 |
| 11:58 |
12:07 I don't understand the MISRA C++ pointer rules: they are
miserable!
17:56 Language subsetting doesn't work.
18:15 Write high-level code via libraries using all the language
features.
19:10 After you build the superset, i.e. language + libraries, you
subset.
19:30 What we want is "C++ on steroids".
23:40 Tool support, i.e. static analysis, is essential.
25:10 No resource leaks: RAII, i.e. no naked new and delete!
32:50 GSL: owner
33:15 Owner
37:00 Besides dangling pointers, there are other pointer misuses:
range errors: fix it with array_view
nullptr dereferencing: fix it with non_null
37:18 We want to minimize run-time checks, because we want to run
fast!
37:50 My least favorite interface:
void f(int *p, int n);
why?
So common!
So error-prone!
Fix it with array_view
40:15 nullptr problems: fix it with not_null
41:30 not_null
42:20 "Smart pointers" are overused to overcome ownership and
dangling pointer problems.
43:12 As the dangling pointers are eliminated by tools, you don't
need to use them!
47:30 Too many rules: the tools should know the rules and will
point you the relevant ones.
49:15 There are too many rules and it doesn't work me! I want lots
of rules.
50:00 I hate camel case.
51:20 We aren't unambitious:
Aim type and resource safety.
Consider all 4+ millions C++ programmers.
Zero-overhead principle.
51:50 We aim to change the way we write code.
53:52 Help wanted! There are currently only two editors: me and
Herb Sutter.
55:02 The basic C++ model is now complete.
57:20 Starts to answer the questions.
No comments:
Post a Comment