| Commit message (Collapse) | Author | Age |
|---|
| |\
| |
| |
| | |
[dgit --quilt=gbp]
|
| | |\
| | |
| | |
| | | |
[dgit --quilt=gbp]
|
| | | |\
| | | |
| | | |
| | | | |
[dgit --quilt=gbp]
|
| | | | |\
| | | | |
| | | | |
| | | | | |
[dgit --quilt=gbp]
|
| | | | | |\
| | | | | |
| | | | | |
| | | | | | |
[dgit --quilt=gbp]
|
| | | | | | |
| | | | | |
| | | | | | |
Gbp-Pq: Name 20no-stray-libz-link.patch
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
t/local/07_sslecho.t.
Fix for GH-160 seems to require SSL_shutdown() calls which were missing from
t/local/07_sslecho.t and Net::SSLeay::sslcat()
These seem to relate to this change in OpenSSL
'Detect EOF while reading in libssl':
https://github.com/openssl/openssl/commit/db943f43a60d1b5b1277e4b5317e8f288e7a0a3a
The newly added OpenSSL error message 'unexpected eof while reading' was seen
before the SSL_shutdown() calls were added.
Bug: https://github.com/radiator-software/p5-net-ssleay/issues/160
Bug-Debian: https://bugs.debian.org/954401
Gbp-Pq: Name openssl-1.1.1e.patch
|
| | | | | | |
| | | | | |
| | | | | | |
Gbp-Pq: Name 20no-stray-libz-link.patch
|
| | | | | | |
| | | | | |
| | | | | | |
Gbp-Pq: Name 20no-stray-libz-link.patch
|
| | | | | | |
| | | | | |
| | | | | | |
Gbp-Pq: Name 20no-stray-libz-link.patch
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Origin: ubuntu
Bug-Debian: https://bugs.debian.org/1010795
Bug-Ubuntu: https://launchpad.net/bugs/1972795
Forwarded: not-needed, only makes sense in a Debian-ish context
Reviewed-by: gregor herrmann <gregoa@debian.org>
Last-Update: 2022-05-10
The test suite checks that the OPENSSL_version_*() calls match the compile-time
constants, which is reasonable in a unit test that is run in the same time
frame as the compilation, as is usually the case. However, this check doesn't
make as much sense in the autopkgtest context, where the src:openssl package
might have been updated for a bugfix release, for instance. Such updates would
trigger an autopkgtest run, and barring regressions in OpenSSL the tests for
this package should reasonably not fail, especially given the explicit
versioning policy from 3.0.0 forward:
https://www.openssl.org/policies/general/versioning-policy.html
Gbp-Pq: Name version_check_on_autopkgtest.patch
|
| | | | | | |
| | | | | |
| | | | | | |
Gbp-Pq: Name 20no-stray-libz-link.patch
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Origin: ubuntu
Bug-Debian: https://bugs.debian.org/1010795
Bug-Ubuntu: https://launchpad.net/bugs/1972795
Forwarded: not-needed, only makes sense in a Debian-ish context
Reviewed-by: gregor herrmann <gregoa@debian.org>
Last-Update: 2022-05-10
The test suite checks that the OPENSSL_version_*() calls match the compile-time
constants, which is reasonable in a unit test that is run in the same time
frame as the compilation, as is usually the case. However, this check doesn't
make as much sense in the autopkgtest context, where the src:openssl package
might have been updated for a bugfix release, for instance. Such updates would
trigger an autopkgtest run, and barring regressions in OpenSSL the tests for
this package should reasonably not fail, especially given the explicit
versioning policy from 3.0.0 forward:
https://www.openssl.org/policies/general/versioning-policy.html
Gbp-Pq: Name version_check_on_autopkgtest.patch
|
| | | | | | |
| | | | | |
| | | | | | |
Gbp-Pq: Name 20no-stray-libz-link.patch
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | | |
Gbp-Dch: Ignore
|
| | | | | | | |
|
| | | | | | | |
|
| | | | | | | |
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | | |
Gbp-Dch: Ignore
|
| |\ \ \ \ \ \
| |/ / / / /
|/| | | | |
| | | | | | |
Update to upstream version '1.94'
with Debian dir 2735366ad965afeb30f6d86b18a37020b56fe6f7
|
| | |\ \ \ \ \ |
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Bump all version numbers from 1.93_05 to 1.94, and summarise major
changes since 1.92.
Closes #460.
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Bump all version numbers from 1.93_04 to 1.93_05, and associate all
reported changes since 1.93_04 with version 1.93_05.
Closes #467.
|
| | | |\ \ \ \ \
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Fixes #465.
|
| | | |/ / / / /
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
This reverts da6292e138644aa1cdd234114c6198675611693c.
The automatic detection of libssl/libcrypto with pkg-config conflicts
with the OPENSSL_PREFIX method of specifying where Makefile.PL should
look for libssl/libcrypto. The one known failure case at the moment is
when OPENSSL_PREFIX points to a copy of OpenSSL that has compression
enabled, but pkg-config points to a copy of OpenSSL that doesn't have
compression enabled - this causes "-lz" to be omitted from the list of
options passed to the linker, causing a compilation failure. There may
be other failure cases.
The automatic detection of libssl/libcrypto is a very useful feature, so
we'll find and fix all the edge cases and bring it back in time for
Net-SSLeay 1.96.
Fixes #465.
|
| | | |\ \ \ \ \
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Closes #462.
|
| | | |/ / / / /
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Bump all version numbers from 1.93_03 to 1.93_04, and associate all
reported changes since 1.93_03 with version 1.93_04.
Closes #462.
|
| | | |\ \ \ \ \ |
|
| | | |/ / / / /
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Tuure's last contribution was the configuration for Travis and AppVeyor, in
2018. Update the maintainership notices and copyright dates in each
module/script to reflect this.
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Most of the exposed API functions were made available in the recent LibreSSL
versions but some had been overlooked for unknown reasons.
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Update Makefile.PL to define OPENSSL_API_COMPAT=908 to suppress compilation
warnings for OpenSSL API functions deprecated in versions older than 0.9.8. The
purpose of this is to turn off all deprecation warnings.
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Bump all version numbers from 1.93_02 to 1.93_03, and associate all
reported changes since 1.93_02 with version 1.93_03.
|
| | | |\ \ \ \ \
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
GH-454 Makefile.PL, XS and installation fixes
Collect a number of updates from different sources.
|
| | | | | | | | | |
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
header files in a non-standard location.
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
With Linux and macOS, for example, compiling OpenSSL and Net::SSLeay with gcc
and clang appears to work these days. It's even possible to compile Net::SSLeay
with a C++ compiler and link it against OpenSSL compiled with a C compiler.
On some other platforms it may still be required to use matching compiler and
flags, but it seems not to be universally required anymore.
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
The current versions of Shining Light OpenSSL MSI installer use installation
paths that are not in the list of prefixes tested by Makefile.PL. At the time
of writing, the MSI installers are Win64OpenSSL-1_1_1w.msi,
Win64OpenSSL-3_0_12.msi, Win64OpenSSL-3_1_4.msi and Win64OpenSSL-3_2_0.msi. The
32bit versions are named as Win32OpenSSL-1_1_1w.msi, etc.
The exe installers seem to use the same locations as the msi installers.
Caveats: static linking seems to require linking against libcrypto_static and
libssl_static. The existing static compilation flag in Makefile.PL currently
searches for libraries with a different path, instead of setting up linking to
against specially named libraries. In other words, static linking seems to have
changed and may change in the future. If you need static linking, you likely
need to modify Makefile.PL to suit your needs.
When both 32bit and 64bit Shining Light OpenSSL versions are installed,
Makefile.PL prefix guessing can choose wrong prefix. In this case it's
currently easiser to set OPENSSL_PREFIX environment variable.
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
-dates to day 1 commit cb98af3caf3b63d7125352aa13d18748ab20ee3a
Fixed some compiler warnings. Courtesy kmx.
Fixed a problem with Win32 detection. Courtesy kmx.
git-svn-id: r334
Committer: Mike McCauley <mikem@airspayce.com> 2012-04-02 04:38:39
-----------------
this caused
-----------------
C:\sources\p5-net-ssleay>perl Makefile.PL
Do you want to run external tests?
These tests *will* *fail* if you do not have network connectivity. [n] y
*** Found OpenSSL-1.1.1g installed in C:\Program Files\OpenSSL-Win32
*** Be sure to use the same compiler and options to compile your OpenSSL, perl,
and Net::SSLeay. Mixing and matching compilers is not supported.
Warning (mostly harmless): No library found for -llibeay32
Warning (mostly harmless): No library found for -lssleay32
Generating a gmake-style Makefile
Writing Makefile for Net::SSLeay
-------------------
With Shining Light precompiled OpenSSL and eventual there is a link failure
since the lib file paths dont exist that are given to EUMM and EUMM
eventually puts in MKF then fed to link.exe cmd line, since the $dir loop
kept spinning and push new paths into %opts even though in an earlier
iteration the .lib file was found on disk
-fix a CC warning too about redefinition, perl core provides this flag
automatic nowadays if the VC version is new enough to need this
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Fix additional cases with code before variable declarations. These were added
before the longer commit by bulk88 which fixed a lot of previously existing
cases.
Note that Perl 5.35.5 and later relax this requirement because 5.35.5 and later
versions use some of C99 (C standard ISO/IEC 9899:1999) features.
Gcc and Clang flag -Werror=declaration-after-statement is especially mentioned
as something that should only be used with Perl versions earlier than 5.35.5.
https://perldoc.perl.org/5.35.5/perldelta#Configuration-and-Compilation
SSLeay.xs currently avoids code before declarations.
|
| | | |/ / / / /
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Cherry-picked with conflict resolution by hvn from PR 188 commit
4ecea059353d0681fac0242c95c22cd8a973501f
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
* Support finding OpenSSL libraries using ExtUtils::PkgConfig
If ExtUtils::PkgConfig is available and there is a pkg-config file
for openssl, use the libraries specified rather than assuming which
libraries need to be linked against. This requires ExtUtils::PkgConfig
version 1.16 or later.
This may be useful for example in the case where OpenSSL is built without
compression support and there may be no libz.so to link against.
In the absence of a recent enough version of ExtUtils::PkgConfig, the
Makefile.PL falls back to the existing configuration method, so adding
this enhancement should be harmless.
* By hvn: Update Changes for ExtUtils::PkgConfig PR 127.
---------
Co-authored-by: Heikki Vatiainen <hvn@radiatorsoftware.com>
|
| | | |\ \ \ \ \
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Expose OpenSSL initialisation and deinitialisation functions OPENSSL_init_crypto and OPENSSL_init_ssl and their related functions. These functions were first added in OpenSSL 1.1.0 and are available when the automatic default library initialisation needs to be overridden.
Starting with Openssl 3.0.0 it seems that Net::SSLeay test 62_threads-ctx_new-deadlock.t needs to override atexit() handler to avoid crashes that happen after the test has successfully finished and the process is exiting. This is now done with OPENSSL_init_crypto(OPENSSL_INIT_NO_ATEXIT). GitHub issue GH-452 lists additional ways to work around the atexit problem.
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
OPENSSL_INIT_NO_ATEXIT was added in OpenSSL 1.1.1b, therefore we need to check
for its presence rather than OPENSSL_init_crypto that was added during OpenSSL
1.1.0 development.
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Fix the #if compilation guard to use correct OpenSSL version.
SSL_SESSION_get0_cipher is added by OpenSSL 1.1.0.
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
OpenSSL 3.0 added atexit() handler that appears to cause crashes with test
62_threads-ctx_new-deadlock.t.
The test is now updated to initialise OpenSSL libcrypto so that atexit()
handler is not called. As a result, OpenSSL libraries are no longer cleaned up
on exit. This is fine for the test since the process that runs it is already
exiting and the test requires no special cleaning up.
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Expose multiple functions related to OpenSSL initialisation and
deinitialisation.
Add new test file 23_openssl_init.t for the newly exposed functions.
|
| | | |/ / / / / |
|
| | | |\ \ \ \ \
| | | | | | | |
| | | | | | | | |
GH-425 and GH-438 Update SSLeay.xs to support C++ compilers. This merge collects a number of updates done during 2023.
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
'extern "C" {}' block has wrapped Perl XS standard includes since the
beginning. The earliest version found from the net is Net_SSLeay.pm version
0.4 in which the following was already present:
#ifdef __cplusplus
extern "C" {
#endif
#include "EXTERN.h"
#include "perl.h"
#include "XSUB.h"
#ifdef __cplusplus
}
#endif
This has been carried over since then, until the recent SSLeay.xs updates to
allow compilation with a C++ compiler caused a compilation failure.
Recent GCC and Clang C++ compilers do not care about the 'extern "C"' wrapper
but Visual Studio Community 2022 flagged the following error (see GH-438):
C:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\um\wspiapi.h(53):
error C2894: templates cannot be declared to have 'C ' linkage
When 'extern "C" {}' is removed completely, the compilation succeeds. This
makes sense since apparently the purpose of 'extern "C" {}' is to wrap one's
own functions instead of headers included from elsewhere. Those headers should
have their own 'extern "C" {}' definitions, if any are needed.
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
'char *.
In C string literals do not have the 'const' qualifier but in C++ they do.
Avoid C++ compiler warnings by changing our TRACE() funciton definition to use
'const char *' instead of plain 'char *'. This is a useful change for C too
because they string passed to the function can be expected to be read-only.
|
