This partially reverts commit f4f2298ef3635acd031cc2ee0e71026cdcda5864
because it caused the compatibility code to call initialization routines
redundantly, leading to memory leakage with OpenSSL 1.1 and broken curl
test-suite in Fedora:
88 bytes in 1 blocks are definitely lost in loss record 5 of 8
at 0x4C2DB8D: malloc (vg_replace_malloc.c:299)
by 0x72C607D: CRYPTO_zalloc (mem.c:100)
by 0x72A2480: EVP_CIPHER_meth_new (cmeth_lib.c:18)
by 0x4E5A550: make_ctr_evp.isra.0 (openssl.c:407)
by 0x4E5A8E8: _libssh2_init_aes_ctr (openssl.c:471)
by 0x4E5BB5A: libssh2_init (global.c:49)
... previously it would default to OpenSSL even with the --with-[crypto]
options used unless you specificly disabled OpenSSL. Now, enabling another
backend will automatically disable OpenSSL if the other one is found.
This introduces a test suite for libssh2. It runs OpenSSH in a Docker
container because that works well on Windows (via docker-machine) as
well as Linux. Presumably it works on Mac too with docker-machine, but
I've not tested that.
Because the test suite is docker-machine aware, you can also run it
against a cloud provider, for more realistic network testing, by setting
your cloud provider as your active docker machine. The Appveyor CI setup
in this commit does that because Appveyor doesn't support docker
locally.
Primarily this is handling cases where top-level files moved into
the docs/ directory. I also corrected a typo and removed the
claim that libssh2 is public domain.
This gets us large file support, is available on any VMS release
in the last decade and more, and gives stat other modern features
such as 64-bit ino_t.
VMS does have stdlib.h, gettimeofday(), and OpenSSL. The latter
is appropriate to hard-wire in the configuration because it's
installed by default as part of the base operating system and
there is currently no libgcrypt port.
%z is a C99-ism that VMS doesn't currently have; even though the
compiler is C99-compliant, the library isn't quite. The off_t used
for the st_size element of the stat can be 32-bit or 64-bit, so
detect what we've got and pick a format accordingly.
Somehow it got Windows-style CRLF endings so convert to just LF,
for consistency as well as not to confuse tools that will regard
the \r as content (e.g. the OpenVMS help librarian).
This commit adds a simple check to see if the offset of the read
request matches the expected file offset.
We could try to recover, from this condition at some point in the future.
Right now it is better to return an error instead of corrupted data.
