Saves comitters' daily life by avoid #include-ing everything from
internal.h to make each file do so instead. This would significantly
speed up incremental builds.
We take the following inclusion order in this changeset:
1. "ruby/config.h", where _GNU_SOURCE is defined (must be the very
first thing among everything).
2. RUBY_EXTCONF_H if any.
3. Standard C headers, sorted alphabetically.
4. Other system headers, maybe guarded by #ifdef
5. Everything else, sorted alphabetically.
Exceptions are those win32-related headers, which tend not be self-
containing (headers have inclusion order dependencies).
`find_time_t` did not work correctly for year older than the Epoch
because it used C's integer division (which rounds negative to zero).
For example, `TIme.new(1933)` returned a wrong time whose year is 1922
in Asia/Kuala_Lumpur because there is no 00:00:00 1st Jan. 1933 in the
time zone.
```
$ TZ=Asia/Kuala_Lumpur ruby -e 'p Time.new(1933)'
1932-12-31 00:00:00 +0700
```
This change fixes the issue by using `DIV` macro instead of `/`.
Now `Time.new(1933)` returns a time in 1933.
```
$ TZ=Asia/Kuala_Lumpur ruby -e 'p Time.new(1933)'
1933-01-01 00:20:00 +0720
```
[Bug #16159]
Akatsuki reported ENV['TZ'] = 'UTC' improved 7x-8x faster on following code.
t = Time.now; 100000.times { Time.new(2019) }; Time.now - t
https://hackerslab.aktsk.jp/2019/12/01/141551
commit 4bc1669127(reduce tzset) dramatically improved this situation. But still,
TZ=UTC is faster than default.
This patch removs unnecessary tzset() call completely.
Performance check
----------------------
test program: t = Time.now; 100000.times { Time.new(2019) }; Time.now - t
before: 0.387sec
before(w/ TZ): 0.197sec
after: 0.162sec
after(w/ TZ): 0.165sec
OK. Now, Time creation 2x faster *and* TZ=UTC doesn't improve anything.
We can forget this hack completely. :)
Side note:
This patch slightly changes Time.new(t) behavior implicitly. Before this patch, it might changes
default timezone implicitly. But after this patch, it doesn't. You need to reset TZ
(I mean ENV['TZ'] = nil) explicitly.
But I don't think this is big impact. Don't try to change /etc/localtime on runtime.
Side note2: following test might be useful for testing "ENV['TZ'] = nil".
-----------------------------------------
% cat <<'End' | sudo sh -s
rm -f /etc/localtime-; cp -a /etc/localtime /etc/localtime-
rm /etc/localtime; ln -s /usr/share/zoneinfo/Asia/Tokyo /etc/localtime
./ruby -e '
p Time.new(2000).zone # JST
File.unlink("/etc/localtime"); File.symlink("/usr/share/zoneinfo/America/Los_Angeles", "/etc/localtime")
p Time.new(2000).zone # JST (ruby does not follow /etc/localtime modification automatically)
ENV["TZ"] = nil
p Time.new(2000).zone # PST (ruby detect /etc/localtime modification)
'
rm /etc/localtime; cp -a /etc/localtime- /etc/localtime; rm /etc/localtime-
End
This removes the related tests, and puts the related specs behind
version guards. This affects all code in lib, including some
libraries that may want to support older versions of Ruby.
WIDEVALUE differs from VALUE in 32bit platform, but some codes assume
that they are the same.
There is `#define STRUCT_WIDEVAL` mode to check the consistency.
This change allows to build with STRUCT_WIDEVAL.
Addrinfo.getaddrinfo and .foreach now accepts :timeout in seconds as
a keyword argument. If getaddrinfo_a(3) is available, the timeout will be
applied for name resolution. Otherwise, it will be ignored.
Socket.tcp accepts :resolv_timeout to use this feature.
This commit is retry of 6382f5cc91.
Test was failed on Solaris machines which don't have "http" in
/etc/services. In this commit, use "ssh" instead.
Addrinfo.getaddrinfo and .foreach now accepts :timeout in seconds as
a keyword argument. If getaddrinfo_a(3) is available, the timeout will be
applied for name resolution. Otherwise, it will be ignored.
Socket.tcp accepts :resolv_timeout to use this feature.
After 5e86b005c0, I now think ANYARGS is
dangerous and should be extinct. This commit deletes ANYARGS from
rb_rescue / rb_rescue2, which revealed many arity / type mismatches.
Build dumped string from base packed data and extended year at
once. Although currently ruby_marshal_write_long() never writes
more than 5 bytes per its format specification, allocate
`sizeof(long)+1` for the sanitation.
Nobuyoshi Nakada
卜部昌平
zverok
प्रथमेश Sonpatki
Tanaka Akira
John Hawthorn
Marcus Stollsteimer
Yusuke Endoh
KOSAKI Motohiro
Kazuhiro NISHIYAMA
Jeremy Evans
NARUSE, Yui
Masaki Matsushita
git
manga_osyo
nobu