Use Boost.Charconv

[grisumbras]

1. Removes embedded older version of Charconv and instead uses the library proper. Fix Remove GCC<5 stuff after Boost 1.87 #1032.
2. Removes embedded version of ryu and uses Charconv for double formatting.

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.23%. Comparing base (7f0bceb) to head (be9028c).

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #993      +/-   ##
===========================================
+ Coverage    93.70%   99.23%   +5.52%     
===========================================
  Files           91       78      -13     
  Lines         9139     8052    -1087     
===========================================
- Hits          8564     7990     -574     
+ Misses         575       62     -513     

Files with missing lines	Coverage Δ
include/boost/json/basic_parser_impl.hpp	98.50% <100.00%> (ø)

... and 1 file with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7f0bceb...be9028c. Read the comment docs.

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[sdarwin]

This is crashing the s390 runners. out-of-memory.
The PR was opened 10 months ago. Have you rebased it on develop?
I now adjusted those runners from 2 jobs to 1 job. Try it again.

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[grisumbras]

I did rebase it on develop.

It is quite interesting that the build is crashing s390 runners. This might be a problem with Boost.Charconv.

[cppalliance-bot]

An automated preview of the documentation is available at https://993.json.prtest2.cppalliance.org/libs/json/doc/html/index.html

[cppalliance-bot]

Benchmark test results. More info at https://benchmark.cppalliance.org/jsonbenchmarks-pullrequests/993/pullrequest.html

[sdarwin]

This might be a problem with Boost.Charconv.

It used more than 8GB of memory.

I have just kicked off another run on 'develop', to compare that.

[sdarwin]

Depending on the test, even the develop branch may get to nearly 8GB.
Charconv is not the cause.
I don't think JSON was using that much memory a couple years ago.
In any case the runner is now parallel-jobs=1 so more memory is available.

[grisumbras]

I've recorded memory usage when building and running tests. The outlier is pretty much compiling parse_into.cpp:

Config	Max mem usage
toolset=gcc variant=release cxxstd=20	3.22GB
toolset=gcc variant=release cxxstd=17	3.15GB
toolset=gcc variant=debug cxxstd=20	3.10GB
toolset=gcc variant=debug cxxstd=17	2.95GB
toolset=clang variant=debug cxxstd=20	2.02GB
toolset=clang variant=debug cxxstd=17	2.00GB
toolset=clang variant=release cxxstd=20	1.51GB
toolset=clang variant=release cxxstd=17	1.42GB

Everthing else is below 0.9GB (https://gist.github.com/grisumbras/5985960c7781c4f657a93605d9e2b0ed).

Unfortuantely, this is inherent to parse_into.cpp. It's very template-heavy. I think, the way we should deal with this is to move it to a separate build target, and run b2 twice: one for regular jobs with regular job count limit, and then the heavy tests one job at a time.

[grisumbras]

BTW, for posterity, the info was collected thusly:

# user-config.jam
using gcc : 13 : /usr/bin/time "-f%C|%M" -o/home/grisumbras/dev/stats -a g++-13 ;
using clang : 20 : /usr/bin/time "-f%C|%M" -o/home/grisumbras/dev/stats -a clang++-20 ;

And then b2 was invoked with

rm /home/grisumbras/dev/stats
b2 test example toolset=gcc,clang variant=debug,release cxxstd=17,20 warnings=extra testing.launcher="/usr/bin/time -f'%C|%M' -o/home/grisumbras/dev/stats -a"

After that the file /home/grisumbras/dev/stats contained collected statistics.