Well, I wrote a program...

#!/usr/bin/perl

use 5.026;
use warnings qw(all);
use IPC::Run3;
use Heap::Fibonacci;
use File::Find;

my $heap      = Heap::Fibonacci->new;
my $processed = 0;

eval {
	local $SIG{INT} = sub { die "SIGINT received" };
	find(
		sub {
			-f or return;

			my ($text, $err);
			run3 [qw(man -Tutf8), $File::Find::name], \undef, \$text, \$err;
			warn "man failed on $_: $?, stderr: $err" if $?;

			my $count = ($text =~ tr/\n//);

			$heap->add(MPLen->new($File::Find::name, $count));
			my $top = $heap->top->val;

It is... taking a long time here.

12700 processed, top 31572 (/usr/share/man/man3/Image::ExifTool::TagNames.3pm.gz)

Ah. Seems I've hit a man bug... man -Tutf8 /usr/share/man/ja/man5/apt_preferences.5.gz seems to loop printing out blank lines forever...

So... ignoring the buggy infinite-length ones:

# 8:   31572  /usr/share/man/man3/Image::ExifTool::TagNames.3pm.gz
# 9:   30588  /usr/share/man/man1/ffmpeg-all.1.gz
#10:   29719  /usr/share/man/man1/ffserver-all.1.gz
#11:   24677  /usr/share/man/man1/ffprobe-all.1.gz
#12:   24383  /usr/share/man/man1/ffplay-all.1.gz
#13:   22264  /usr/share/man/man1/c++.1.gz
#14:   22264  /usr/share/man/man1/gcc-7.1.gz
#15:   22264  /usr/share/man/man1/gcc.1.gz
#16:   22264  /usr/share/man/man1/cc.1.gz
#17:   22264  /usr/share/man/man1/g++.1.gz
#18:   22264  /usr/share/man/man1/g++-7.1.gz

(Those start at #8 due to the bug.)

Sent a bug report to Debian. Well, it'll show up at man-db eventually.. currently being greylisted.

@Gilles that's an increasingly common problem - lots of stuff only has online docs these days. even worse, lots of stuff doesn't even have a man page or any documentation at all (or, just as useless but even more annoying, a man page with a single-line giving a useless summary of the command "foo-wrangler wrangles foos")

so I'm browsing markdown files that automatically convert to clickable usable links in my browser.

is there a way to get lynx to automatically detect http:// urls?

I'm using strapdown.js

strapdownjs.com

@cas Sometimes Debian does package the online stuff. But I don't know if they do it consistently or for Python. But it's true - either you have to regularly buy programming books (and even those might not be sufficient), or you need a good net connection.

@FaheemMitha python-doc ;-)

But at least Python has documentation; lots of things don't.

@FaheemMitha right, and “source code as documentation” isn’t nearly sufficient

@StephenKitt Is that everything on the web?

also lots of things have incorrect documentation, which means one ends up needing to read the source too /o\

@StephenKitt Not unless you can read and understand source code in a variety of languages at superspeed.

@FaheemMitha everything on the Python documentation web site, AFAICT

@StephenKitt Or outdated documentation.

@StephenKitt Oh, that's good. I wonder how they package it. Web scraping?

Or does Python have it as part of the source?

@FaheemMitha it’s part of the source

@StephenKitt Oh. That's good.

Debian packages aren’t allowed to pull in external resources at build time, which excludes web scraping in any form

@StephenKitt Oh right, I forgot.

It looks like the C++ Primer is now a bit ouf date - amazon.com/gp/product/0321714113

Published in 2012.

@FaheemMitha there aren’t many C++ books covering C++14 and C++17

even Stroustrup’s The C++ Programming Language stops at C++11

@StephenKitt Does Stroustrup still update his?

@FaheemMitha he did, extensively, for C++11 in 2013

@StephenKitt I see. I didn't know that. But I haven't been keeping track of C++ in recent years. Sounds like you have been.

@FaheemMitha just a little, I don’t do much C++ day-to-day

It looks like C++ is asymptotically approaching the functionality of a reasonable language like Common Lisp. But if and when it approaches such a zenith, it will also be completely incomprehensible.

Around 2050, perhaps.

According to WP, gcc 6 has implemented Concepts even though it's not yet in the C++ standard. Which seems a little surprising.

@FaheemMitha not really, C/C++ compilers often include new features ahead of their adoption in the standard

Concepts already exists as a Technical Specification

@StephenKitt I thought Concepts was a big thing.

Maybe they implement small stuff ahead of time.

Has anyone given it a whirl? gcc 6 is in Debian stable.

@FaheemMitha no, it’s explicitly marked experimental

@StephenKitt I meant the specification.

@FaheemMitha and I imagine some people have, part of the idea of implementing TSs ahead of their adoption is to work out issues by trying things out

I read that there is talk about having it in the C++ 20 standard.

@FaheemMitha oh, then no, it’s not stable

@StephenKitt Oh, I see. You mean that the committee is using gcc as a testbed?

Or people helping the committee?

@FaheemMitha feedback from experimental implementations goes back to the committee, yes

@FaheemMitha there’s probably more incentive to develop “big things” ahead of time, to win the implementation race once the standard is ratified

I think the language has changed quite a lot since the C++ 98 standard. Which was the state of play for a while. At least it seems so.

@StephenKitt Good point.

@FaheemMitha indeed; it’s become so flexible now that two people can write valid C++ in completely different styles, resulting in code that the other can’t understand

which is great for maintenance

@StephenKitt Yes, fun all around.

For a while nothing much seemed to happen. Now there are new standards every few years. I wonder what changed.

Who funds these activities, anyway?

@FaheemMitha a bunch of IT companies: isocpp.org/std/the-committee

@StephenKitt Oh. So a corporate thing. I should have guessed.

although ISO is structured by member country, not employers

@FaheemMitha presumably there are some academics in the WG, but yes, it’s mostly funded by corporations with an interest in the standard

And in the meantime CL has only had one standardization.

@FaheemMitha but it emerged perfectly formed from the deep, so that’s all that will ever be needed

@StephenKitt Heh. I think a revision (maybe not a major one) would not hurt. Sand off some of the rough edges.

Standardise some things that haven't been standardised.

just saw a question that made me realise that all my simple number-matching regexps are wrong. [+-]? *[0-9.]+ is broken in many locales. should be (at minimum) [+-]? *[0-9.,]. obvious in retrospect, and something i knew...but just didn't bother to take into account.

@cas plus thousands separators

but C is the only true locale for parsing

most locales that use , for decimal point use . for thousands separators. some locales use a space, but i really don't want to use a space in a simple number-matcher....if i want to deal with that, i'll google for a complete regex with all the bells & whistles (or use Regexp::Common::Number in perl).

@cas you’ll also find ' in some locales, but yes, either use the C locale or use a proper tool for the job

anyway AFAIAC, a number with thousands separators isn't numeric, it's a descriptive string :)

dammit

I accidentally closed the very question I meant to keep open: stackoverflow.com/questions/49201837/mem-leaks-with-shared-ptr

can I fix it?

I suppose SO could reject the migration?

I apologize, and will return my "trusted user" fake internet trophy immediately

If I sudo vi /etc/hosts and say :sh

I get root access

How come?

@overexchange Because you asked for it?

Asked when? when I say :sh?

where did I ask for root access

?

@overexchange Where you wrote sudo...

/etc/hosts file is a mapping between ip and hostnames

I am in sudoer list to be able to open /etc/hosts

How this is related to getting root access?

@overexchange Everything in the sudoers list is related to getting root access.

sudo apt-get install will not let me get root access

Maybe ask a question on the site