| Age | Commit message (Collapse) | Author |
|---|
|
Use maps in definitions to make them more readable.
|
|
|
|
* goeldeepak/erts/fix_inet_gethost_long/ERL-61/PR-1345/OTP-14310:
This patch fixes the issue in which erlang fails to start if the hostname is 64 characters on a linux system.
|
|
|
|
The above commit erroneously declared 'ppc64le' as a 'ppc64' for
HiPE enabling. These two architectures are different and one can
NOT expect that native code generated for one architecture runs
on the other one. Rather than reverting this commit, make sure
that the architecture is declared correctly instead of pretending
it's a PPC64.
|
|
Fix a bug with the use of the aes_cfb128 cipher by calling the correct
underlying openssl interface function when the cipher is specified.
|
|
Demonstrate a bug with AES CFB 128 for certain key sizes introduced
with the Erlang 19.0 release. The code in the block_crypt_nif function
in the crypto.c source file incorrectly calls aes_cfb_8_crypt when the
specified cipher is aes_cfb8 or aes_cfb128 and the key size is 24 or
32. The aes_cfb_8_crypt function calls the AES_cfb8_encrypt function
from the openssl interface, but this is incorrect when the cipher is
aes_cfb128.
Unfortunately the test cases in the crypto test suite are insufficient
to detect an issue like this because it exercises the encryption and
decryption roundtrip using the same incorrect underlying function. The
problem was observed when trying to update an application to Erlang 19
that attempted to decrypt data that was encrypted using aes_cfb128 by
another source. In this commit I altered the crypto test suite to
provide a demonstration of this problem.
|
|
|
|
* egil/erts/fix-copy-sharing-assert:
erts: Fix faulty assert for refs in copy sharing
|
|
Improve no-pie solution from bug report ERL-294
|
|
|
|
DID_TRAP will read 'ret' even when error is returned.
Found by valgrind.
|
|
Store config when exiting app and restore config when starting again.
|
|
HiPE: Fix ERL-278: Fix range analysis miscompilation bug
|
|
|
|
Previously changing tabs during high cpu-load, could make the
change go unnoticed, and thus the graphs did not get updated.
|
|
* sverker/hipe-long-lived:
Make hipe_bifs:alloc_data/3 pad addr to alignment
erts: Change HIPE allocations from sys_alloc
|
|
* lukas/erts/hide-debug-consoles:
erts: Only show debug consoles if ERL_CONSOLE_MODE is defined
|
|
* lukas/erts/fix_enif_inspect_binary_emasculate/OTP-14304:
erts: Add enif_inspect_binary emasculation of refc bins
|
|
|
|
|
|
Selection was lost after updates
|
|
* lukas/kernel/gen_tcp_close_docs:
kernel: Expand gen_tcp:close docs with send text
|
|
This commit attempts to clarify some of the guarantees given by
the TCP standard when issuing close. This is quite a complex
topic so there are probably corner cases still left, but this
at least tells the user that things can go wrong when doing
a close.
|
|
ERL-379
* sverker/make-config-dep:
erts: Make generated files depend on Makefile
|
|
7814ec18b made hipe_bifs:alloc_data/3 expect alignments greater than 8
from erts_alloc(), which is not something that it guarantees. Fix it up
so that it pads the allocation up to the required alignment, in case it
does not initially satisfy the alignment requirement.
|
|
hipe: Fix alignment of byte-sized constants
|
|
HiPE: Fix off-by-one bug in register allocators
Fix for PR-1380
|
|
HiPE's range analysis would not update the arguments of a callee when
the result of the call was ignored.
Fixes ERL-278.
|
|
|
|
This is a poor man's solution that allows to build and test the
system with all files compiled to native code simply by setting
the ERL_COMPILER_OPTS environment variable. Better solutions,
like automatically setting the no_native option whenever the
compiler sees an on_load attribute, obviously exist but require
more time to implement.
|
|
Emulate active once in such a way that data recived by the TLS
connection process, but not fetch via active once option
by the user, can be delivered at next active once before final close.
|
|
erl_bif_types contains a fixed and improved copy-paste (obvious from the
dead Max_range2_leq_zero if branches) of hipe_icode_range:range_rem/2.
For now, delete the dead code and propagate back fixes and improvements
to hipe_icode_range.
|
|
Remove inets from runtime dependencies in tools
|
|
Smoke test HiPE in the 32-bit build
|
|
Commit d8ff1e3 removed cover_web and therefore
tools no longer depends on inets.
|
|
|
|
|
|
HiPE is tested when running dialyzer in the 64-bit build. To avoid
running out of memory or time, dialyzer is not run in the 32-bit
build job.
Do a smoke test of HiPE by letting dialyzer create a small PLT.
We could catch more bugs by using 'configure --enable-native-libs',
but I am worried that the build would not finish in 50 minutes
(the time limit for a Travis job).
|
|
* egil/compiler/is_tagged_tuple/OTP-12148:
Update primary bootstrap
compiler: Cover beam_record in tests
hipe: Transform is_tagged_tuple instruction
compiler: Add is_tagged_tuple instruction
|
|
|
|
|
|
|
|
Rewrite the instruction stream on tagged tuple tests.
Tagged tuples means a tuple of any arity with an atom as its first element.
Typically records, ok-tuples and error-tuples.
from:
...
{test,is_tuple,Fail,[Src]}.
{test,test_arity,Fail,[Src,Sz]}.
...
{get_tuple_element,Src,0,Dst}.
...
{test,is_eq_exact,Fail,[Dst,Atom]}.
...
to:
...
{test,is_tagged_tuple,Fail,[Src,Sz,Atom]}.
...
|
|
hipe_regalloc_loop considers SpillLimit to be an inclusive lower bound,
the allocators considered it to be an exclusive lower bound. The
allocators are changed to also consider it an inclusive lower bound.
This caused the register allocators to occasionally spill the first
"unspillable" temporary. This caused a failure in a newly added
assertion when hipe-compiling dets_v9 on x86.
|
|
* ingela/ssl/make-cert-test-data/OTP-14294:
ssl, public_key: Add functionality for generating X509 cert test data
|
|
For now this functionality is located in ssl. And existing
public_key function is extended. However some of the functionality may
be moved to public_key in a later stage.
|
|
* ingela/ssl/dtls-frag:
dtls: Handle overlapping fragments
|
|
Fragment reassembling needs to handle that a smaller
fragment then sent originally might overlap an earlier
received fragment.
|
|
Old ssh_benchmark_SUITE was very resource demanding. The new
ssh_bench_SUITE is start of a new way to measure.
|
