vadz
commented
7 years ago So I tried the following patch:
diff --git a/lib/OpenSSL.pm6 b/lib/OpenSSL.pm6
index 7c3d96c..3840f8d 100644
--- a/lib/OpenSSL.pm6
+++ b/lib/OpenSSL.pm6
@@ -35,8 +35,14 @@ method new(Bool :$client = False, ProtocolVersion :$version = -1) {
# can't find the required libeay32.dll anywhere in the path, and so fails to load the dll
OpenSSL::EVP::EVP_aes_128_cbc();
- OpenSSL::SSL::SSL_library_init();
- OpenSSL::SSL::SSL_load_error_strings();
+ try {
+ CATCH {
+ # The mask is OPENSSL_INIT_LOAD_SSL_STRINGS | OPENSSL_INIT_LOAD_CRYPTO_STRINGS.
+ default { OpenSSL::SSL::OPENSSL_init_ssl(0x00300000, Nil); }
+ }
+ OpenSSL::SSL::SSL_library_init();
+ OpenSSL::SSL::SSL_load_error_strings();
+ }
my $method;
given $version {
diff --git a/lib/OpenSSL/SSL.pm6 b/lib/OpenSSL/SSL.pm6
index b3c3e79..4faf6a9 100644
--- a/lib/OpenSSL/SSL.pm6
+++ b/lib/OpenSSL/SSL.pm6
@@ -37,6 +37,7 @@ class SSL is repr('CStruct') {
}
our sub SSL_library_init() is native(&ssl-lib) { ... }
+our sub OPENSSL_init_ssl(uint64, OpaquePointer) is native(&ssl-lib) { ... }
our sub SSL_load_error_strings() is native(&ssl-lib) { ... }
our sub SSL_new(OpenSSL::Ctx::SSL_CTX) returns SSL is native(&ssl-lib) { ... }but, as expected, this is not enough, it now crashes later in t/02-socket.t:
(gdb) r
Starting program: /usr/bin/moar --execname=-zsh --libpath=/usr/share/nqp/lib --libpath=/usr/lib/perl6/runtime --libpath=/usr/share/perl6/lib --libpath=/usr/share/perl6/runtime /usr/share/perl6/runtime/perl6.moarvm -Ilib t/02-socket.t
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
1..4
Program received signal SIGSEGV, Segmentation fault.
0x00007ffff410be30 in SSL_get_rbio () from /usr/lib/x86_64-linux-gnu/libssl.so
(gdb) bt
#0 0x00007ffff410be30 in SSL_get_rbio () from /usr/lib/x86_64-linux-gnu/libssl.so
#1 0x00007ffff410be78 in SSL_set_bio () from /usr/lib/x86_64-linux-gnu/libssl.so
#2 0x00007ffff71fe038 in ffi_call_unix64 () from /usr/lib/x86_64-linux-gnu/libffi.so.6
#3 0x00007ffff71fda9a in ffi_call () from /usr/lib/x86_64-linux-gnu/libffi.so.6
#4 0x00007ffff791e3d0 in MVM_nativecall_invoke () from //usr/lib/moar/libmoar.so
#5 0x00007ffff78f5c4b in MVM_interp_run () from //usr/lib/moar/libmoar.so
#6 0x00007ffff79a0e01 in MVM_vm_run_file () from //usr/lib/moar/libmoar.so
#7 0x00005555555550d8 in ?? ()
#8 0x00007ffff74212b1 in __libc_start_main (main=0x555555554d20, argc=9, argv=0x7fffffffe678, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7fffffffe668) at ../csu/libc-start.c:291
#9 0x000055555555514a in ?? ()I'm afraid I have to abandon now, all I wanted was to write a small script making parallel HTTPS requests and unfortunately this just seems to be impossible currently :-(
Xliff
ugexe
This doesn't seem to work with OpenSSL 1.1.0, as included in Debian Sid, right now: the tests fail when installing it with error messages similar to this:
apparently because
SSL_library_init()has been replaced withOPENSSL_ssl_init()now (at least the former function is not in my/usr/lib/x86_64-linux-gnu/libssl.so.1.1, but the latter is). I am probably going to try changing a local version to use the new function in desperation (because as it stands, HTTP::UserAgent is just completely unusable with HTTPS for me because of #1 under Jessie and this one in Sid.), but I guess this is not going to be enough as there are probably other API changes in 1.1.