| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
* Add bounds checks for allocation multiplication
* Simplify the error handling by adding a flag for whether the strings
were loaded.
|
|
|
|
|
|
* Add error handling for all OpenSSL calls.
|
|
|
|
* Add error handling for all Erlang calls.
|
|
|
|
|
|
|
|
* Add error handling for all OpenSSL calls.
- However, disable custom crypto mem functions for LibreSSL
since it has never supported it.
|
|
* Simplify error checking.
* Add bounds checking before casting.
* Add error checking for all OpenSSL calls.
|
|
* Add error handling for OpenSSL calls.
* Change dlen to signed since BN_num_bytes() returns int.
* Use enif_make_badarg() on error since it only returned
undefined before in one type of error.
|
|
* Add error checking for OpenSSL calls.
* Change dlen from unsigned to signed since BN_num_bytes is int.
* Add bounds checking before casting.
* Consolidate all freeing to one location on error or success.
|
|
* Add error handling for all OpenSSL calls
- There was nothing returned on error before so use enif_make_badarg().
* Add bounds checking before casting.
|
|
* Add bounds checking.
* Add error checking for OpenSSL calls.
* Only set *bnp on success.
|
|
* Add bounds checking.
* Add error checking for OpenSSL calls.
* Only set argument *bnp on success.
|
|
* Bounds check before casting.
|
|
* Add error handling for all OpenSSL calls.
|
|
* Add error checking for all OpenSSL calls.
|
|
* Add bounds check before casting.
|
|
* Add error checking and use enif_make_badarg() on error.
* Use size_t when using sizeof().
* Move initialization away from declaration so it's not as easy to miss.
|
|
* Change the parameter from int to size_t.
- Only caller doesn't need to change since it was already passing sizeof().
* Add unsigned wrapping checks.
|
|
* Simplify logic by having incremental allocation and only free on error on one place.
* Add error checking on all OpenSSL calls.
* Make it explicit when you need to be careful with non-reference counted pointers.
- set0 calls will use the pointer values without ref counting.
- On success, set pointers to NULL to avoid double frees since the struct
is now responsible for freeing the resources.
|
|
* Simplify logic by having incremental allocation and only free on error
in one place.
* Add error checking on all OpenSSL calls.
* Make it explicit when you need to be careful with non-ref counted pointers.
- set0 calls will save the pointer without reference counting.
- On success, set pointers to NULL to avoid double frees since the struct
is now responsible for freeing the resources.
|
|
* Bug fix: ECDH_compute_key was not being checked for failures
- That function returns 0 on failure and never returns < 0.
- Using <= 0 check because OpenSSL uses that internally and on their wiki.
* Remove unnecessary variable i
* Make the gotos always flow downward rather than jumping back.
|
|
* Check return values on all OpenSSL and Erlang calls.
* Remove trailing semicolon after function definition.
|
|
* Check return values on all OpenSSL calls.
* Remove trailing semicolon after function definition.
|
|
* Fix a possible memory leak where enif_alloc_binary() wasn't always
converted into a term or freed in the error path.
|
|
* Bounds check key.size before casting.
|
|
* pkey is only set on success.
|
|
* Add additional error checking.
|
|
* Add error checking on all OpenSSL and Erlang calls.
|
|
A number of files will do bounds checking.
|
|
Make it NULL safe.
|
|
Make it NULL safe.
|
|
Make it NULL safe.
|
|
Make it NULL safe.
|
|
* maint:
Remove unsafe optimization for delaying creation of stackframe
Conflicts:
lib/compiler/src/v3_codegen.erl
|
|
* bjorn/compiler/interim-tuple-bug/OTP-15501/ERL-807:
Remove unsafe optimization for delaying creation of stackframe
|
|
* maint:
ssh: Wait a bit for the docker client to finnish in ssh_compat_SUITE
|
|
Conflicts:
lib/ssl/src/ssl_handshake.erl
|
|
The testing with a remote client in a docker and a local server could
show some hazards. For instance instructing the client to sftp a
file to the server is not necessarily ready in the server's file system
when the common_test is scheduled.
To avoid that the test suite dooms a file operation to not have been
performed, simply sleep at increasing times for a while with testing
if it is done. After a while give up and signal a fault.
|
|
Make all compiler options work in the option list
|
|
* ingela/ssl/error-handling-should-be-throw/OTP-15502:
ssl: Clean up extension handling
|
|
b89044a800c4 introduced an optimization that tries to delay creation
of stack frames. It turns out that this optimization is not always
safe. (See the new test case for an example.)
Since the code generator is completely rewritten in the `master`
branch for the upcoming OTP 22 release, it does not make sense trying
to mend this optimization. It is better to remove it. Out of a sample
of about 1000 modules in OTP, about 50 of them are changed as a result
of removing this optimization.
The compiler in OTP 22 will do the same optimization in a cleaner,
safer, and more effective way.
https://bugs.erlang.org/browse/ERL-807
|
|
* maint:
Fix sftpd interop for SSH_FXP_STAT
|
|
Some SFTP client's have been discover to claim one version of the
protocol and send messages in a different version.
In the spirit of Postel's law, we can for SSH_FXP_STAT, ignore the
differences in the protocol, as we are not doing anything with the flags
anyway.
|
|
* maint:
Updated OTP version
Prepare release
|
