This leads to double requiring "ext/psych/lib/psych/versions.rb" and
".ext/common/psych/versions.rb", which produces doxens of warnings
during `make test-all`.
20200608T123004Z.log.html.gz
```
[15092/20234]
TestRDocOptions#test_init_with_encoding/home/chkbuild/chkbuild/tmp/build/20200608T123004Z/ruby/ext/psych/lib/psych.rb:233: warning: already initialized constant Psych::LIBYAML_VERSION
/home/chkbuild/chkbuild/tmp/build/20200608T123004Z/ruby/.ext/common/psych.rb:233: warning: previous definition of LIBYAML_VERSION was here
/home/chkbuild/chkbuild/tmp/build/20200608T123004Z/ruby/ext/psych/lib/psych.rb:235: warning: already initialized constant Psych::NOT_GIVEN
/home/chkbuild/chkbuild/tmp/build/20200608T123004Z/ruby/.ext/common/psych.rb:235: warning: previous definition of NOT_GIVEN was here
/home/chkbuild/chkbuild/tmp/build/20200608T123004Z/ruby/ext/psych/lib/psych.rb:271: warning: method redefined; discarding old load
/home/chkbuild/chkbuild/tmp/build/20200608T123004Z/ruby/.ext/common/psych.rb:271: warning: previous definition of load was here
...
```
The same issue happened in fiddle.gemspec, and this change applies the
same fix as [fiddle](https://github.com/ruby/fiddle/issues/34).
Removing it triggers the following warnings when running `bundle` under
jruby from the root of the `psych` repo prints the following warnings:
```
/path/to/jruby-9.2.9.0/lib/ruby/gems/shared/gems/psych-3.1.0-java/lib/psych/versions.rb:7: warning: already initialized constant VERSION
/path/to//jruby-9.2.9.0/lib/ruby/gems/shared/gems/psych-3.1.0-java/lib/psych/versions.rb:10: warning: already initialized constant DEFAULT_SNAKEYAML_VERSION
```
This is because bundler loads the versions file relatively from the
local gemspec, and then internally loads the psych gem, causing the
redefinition warnings.
Instead, we modify the $LOAD_PATH so that when working locally on the
`psych` repo, the local version of `psych` gets used.
a3fc8191a7
Current version is not installable because the gemspec is invalid, since
it includes a `.travis.yml` file that no longer exists, so `rake
install` fails like this:
```
$ rake install
rake aborted!
WARNING: See https://guides.rubygems.org/specification-reference/ for help
ERROR: While executing gem ... (Gem::InvalidSpecificationException)
[".travis.yml"] are not files
```
32b18ca7ca
It loads `ext/fiddle/lib/fiddle/version.rb`, which causes constant
redefinition warning:
20200523T153003Z.log.html.gz
```
[ 6317/20193] TestDefaultGems#test_validate_gemspec/home/chkbuild/chkbuild/tmp/build/20200523T153003Z/ruby/ext/fiddle/lib/fiddle/version.rb:2: warning: already initialized constant Fiddle::VERSION
/home/chkbuild/chkbuild/tmp/build/20200523T153003Z/ruby/.ext/common/fiddle/version.rb:2: warning: previous definition of VERSION was here
= 0.16 s
```
This changeset read the version by manual parsing hack which is also
used in stringio and zlib.
* Allow access to a struct's underlying memory with `struct[offset, length]`.
* Make accessing a struct's underlying memory more convenient.
* refactor memory access unit tests for improved clarity
c082c81bb5
Ragel generates a code `0 <= (*p)` where `*p` is char.
As char is unsigned by default on arm and RISC-V, it is warned by gcc:
```
compiling parser.c
parser.c: In function ‘JSON_parse_string’:
parser.c:1566:2: warning: comparison is always true due to limited range of data type [-Wtype-limits]
if ( 0 <= (*p) && (*p) <= 31 )
^
parser.c:1596:2: warning: comparison is always true due to limited range of data type [-Wtype-limits]
if ( 0 <= (*p) && (*p) <= 31 )
^
```
This change removes the warning by substituting the condition with
`0 <= (signed char)(*p)`.
Let's revert the changes for now, as it cannot be included in the 2.2.0
release.
My comment on #257:
> A blocker is OpenSSL::SSL::SSLContext#add_certificate_chain_file. It
> has a pending change and I don't want to include it in an incomplete
> state.
>
> The initial implementation in commit 46e4bdba40c5 was not really
> useful. The issue is described in #305. #309 extended it
> to take the corresponding private key together. However, the new
> implementation was incompatible on Windows and was reverted by #320 to
> the initial one.
>
> (The prerequisite to implement it in) an alternative way is #288, and
> it's still cooking.
This effectively reverts the following commits:
- dacd08937ccd ("ssl: suppress test failure with SSLContext#add_certificate_chain_file", 2020-03-09)
- 46e4bdba40c5 ("Add support for SSL_CTX_use_certificate_chain_file. Fixes #254.", 2019-06-13)
ea925619a9
Implement OpenSSL::PKey::PKey#oid as a wrapper around EVP_PKEY_id().
This allows user code to check the type of a PKey object.
EVP_PKEY can have a pkey type for which we do not provide a dedicated
subclass. In other words, an EVP_PKEY that is not any of {RSA,DSA,DH,EC}
can exist. It is currently not possible to distinguish such a pkey.
Also, implement PKey#inspect to include the key type for convenience.
dafbb1b3e6
Maybe due to e1855100e4, CentOS, RHEL, and
Fedora CIs have started failing with SEGV. Try to avoid
ffi_closure_alloc on those environments.
20200512T183004Z.fail.html.gz
20200512T183004Z.fail.html.gz
20200512T183003Z.fail.html.gz
to define HAVE_FFI_CLOSURE_ALLOC.
The macro is used in closure.c, so have_func check is needed.
If pkg-config is not installed, extconf.rb fails to detect the version
of libffi, and does not add "-DUSE_FFI_CLOSURE_ALLOC=1" even when system
libffi version is >= 3.2.
If USE_FFI_CLOSURE_ALLOC is not defined, closure.c attempts to check if
HAVE_FFI_CLOSURE_ALLOC is defined or not, but have_func was removed with
528a3a1797, so the macro is always not
defined.
This resulted in this deprecation warning:
20200512T123003Z.log.html.gz
```
compiling closure.c
closure.c: In function 'initialize':
closure.c:265:5: warning: 'ffi_prep_closure' is deprecated: use ffi_prep_closure_loc instead [-Wdeprecated-declarations]
265 | result = ffi_prep_closure(pcl, cif, callback, (void *)self);
| ^~~~~~
In file included from ./fiddle.h:42,
from closure.c:1:
/usr/include/x86_64-linux-gnu/ffi.h:334:1: note: declared here
334 | ffi_prep_closure (ffi_closure*,
| ^~~~~~~~~~~~~~~~
```
https://gcc.gnu.org/gcc-10/changes.html#c
> * GCC now defaults to `-fno-common`. As a result, global
> variable accesses are more efficient on various targets. In
> C, global variables with multiple tentative definitions now
> result in linker errors. With `-fcommon` such definitions are
> silently merged during linking.
