Now that we can become "in init" directly after the call, test the
various scenarios where explicit SSL_do_handshake() calls can come
into play.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14817)
The initial implementation always deferred the generation of the
requested ticket(s) until the next application write, but this
is not a great fit for what it actually does, architecturally wise.
A request to send a session ticket means entering back into the
handshake state machine (or "in init", as it's known in the
implementation). The state machine transition is not something that
only occurs at an application-data write, and in general could occur at
any time. The only constraint is that we can't enter "init" while in
the middle of writing application data. In such cases we will need to
wait until the next TLS record boundary to enter the state machine,
as is currently done.
However, there is no reason why we cannot enter the handshake state
machine immediately in SSL_new_session_ticket() if there are no
application writes pending. Doing so provides a cleaner API surface to
the application, as then calling SSL_do_handshake() suffices to drive
the actual ticket generation. In the previous state of affairs a dummy
zero-length SSL_write() would be needed to trigger the ticket
generation, which is a logical mismatch in the type of operation being
performed.
This commit should only change whether SSL_do_handshake() vs zero-length
SSL_write() is needed to immediately generate a ticket after the
SSL_new_session_ticket() call -- the default behavior is still to defer
the actual write until there is other application data to write, unless
the application requests otherwise.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14817)
apps/progs.pl will have apps/progs.c as output, and on some systems,
the output file of a program is locked against reading.
Unfortunately, apps/progs.c is also part of the sources that make up
apps/openssl, so it's necessary to mark that file in a way that makes
progs.pl skip over it.
Fortunately, this is easily done with a special attribute in
apps/build.info and a simple adaptation of apps/progs.pl.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15332)
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15316)
Current VMS C-RTL does not have <sys/select.h>. <sys/socket.h> is
a good enough replacement to get fd_set.
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15319)
It would try to define OPENSSL_SYS_VMS if that macro is defined.
That's just not right.
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15319)
We have pretty long symbol names, so they need to be shortened to fit
in the linker's 31 character limit on symbols.
Symbol name shortening with the VMS C compiler works in such a way
that a symbol name that's longer than 31 characters is mangled into
its first original 22 characters, followed by a dollar sign and the
32-bit CRC of the original symbol name in hexadecimal.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15317)
crypto/ec/curve448/ has a series of inclusions that throws VMS C
off, so we compensate for it the same way as we have done before.
Fixes#14247
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15317)
Instead of what we used to do, put all include directories in a number
of DCL variables and generate the /INCLUDE qualifier value on the
command line, we instead generate VMS C specific header files with
include directory pragmas, to be used with the VMS C's /FIRST_INCLUDE
qualifier. This also shortens the command line, the size of which is
limited.
VMS C needs to have those include directories specified in a Unix
form, to be able to safely merge #include paths with them when
searching through them.
Fixes#14247
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15317)
`use lib` needs Unix formatted paths. For VMS, it means that we must
make sure to convert paths, and we may as well generalise it.
In this case, we need to adapt the functions sourcedir() and sourcefile()
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15317)
It was looking in the wrong place in %unified_info to determine if the
library would be installed or not.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15317)
VMS unpackers will typically convert any period ('.') in directory
names to underscores, since the period is a path separator on VMS,
just like '/' is a path separator on Unix. Our fallback mechanism
needs to account for that.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15317)
Get it back in sync with the other templates, and correct a few syntax
errors that have crept in.
Fixes#14247
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15317)
VMS C can be notoriously informative about certain things, such as
unsupported pragmas. The case here is that it doesn't support
"#pragma once", and since we use those quite a lot, that's a lot of
repeated information. We simply turn that warning off.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15317)
The memory footprint of how we produced the Makefile was quite...
important, because we have all the processing in one perl snippet, and
generate the details of the build file by appending to the "magic"
variable $OUT. The result is that this variable gets to hold the
majority of the build file text, and depending on memory reallocation
strategies for strings, the heap may hold multiple (possibly not just
a few) copies of this string, almost all of them "freed" but still
taking up space. This has resulted in memory exhaustion.
We therefore change strategy, and generate the build file in two
phases, where the first phase generates the full template using small
perl snippets for each detail, and the second phase processes this
template. This is much kinder to process memory.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15310)
quotify1() and quotify_l() were in OpenSSL::Template, but should be
more widely usable.
configdata.pm.in's out_item() is also more widely useful and is
therefore moved to OpenSSL::Util as well, and renamed to dump_data().
Reviewed-by: Tomas Mraz <tomas@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15310)
No longer needed after rewrite of cmd-nits
Reviewed-by: David von Oheimb <david.von.oheimb@siemens.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15311)
If you don't have the base or default providers loaded and therefore there
are no encoders/decoders or store loaders then the error messages can be
cryptic. We provide better hints about how to fix the problem.
Fixes#13798
Reviewed-by: Paul Dale <pauli@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/15306)
