| Age | Commit message (Collapse) | Author |
|---|
|
This is possible now that we track types on a per-value basis, and
no longer need to care whether the source tuple's register has been
clobbered by the time we infer the type.
|
|
|
|
This is a rather subtle but important distinction. While tracking
types on a per-register basis is fairly effective, it forces us to
track which registers alias each other, and makes it tricky to infer
types over large blocks of code as instruction arguments may have
been clobbered between definition and inference.
Tracking types on a per-value basis makes us immune to these
problems.
|
|
* maint:
edoc: Print a helpful message if reading source file fails
|
|
* hasse/edoc/helpful_message/OTP-15605/ERL-841:
edoc: Print a helpful message if reading source file fails
|
|
The stronger compiler optimizations made the test case fail.
|
|
5239eb0c62a9 removed some optimizations in `sys_core_fold`, and
because of that two warnings are no longer emitted.
|
|
This reverts commit 5c7787077e24b0da74cce248a91fcdb7e0d7bb3f.
|
|
This reverts commit df130102cdeca8d35fec95a0c926fd1cfec54eab.
|
|
|
|
|
|
I have no idea how this escaped us for so long...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Consider the following code:
bme(Int) ->
TagInt = Int band 2#111,
Tag = case TagInt of
0 -> a; 1 -> b; 2 -> c; 3 -> d;
4 -> e; 5 -> f; 6 -> g; 7 -> h
end,
case Tag of
g -> expects_g(TagInt, Tag);
h -> expects_h(TagInt, Tag);
_ -> Tag = id(Tag), ok
end.
expects_g(6, Atom) -> Atom = id(g), ok.
expects_h(7, Atom) -> Atom = id(h), ok.
The type optimization pass would recognize that TagInt can only be
[0 .. 7], so the first 'case' would select_val over [0 .. 6] and swap
out the fail label with the block for 7.
A later optimization would merge this block with 'expects_h' in the
second case, as the latter is only reachable from the former.
... but this broke down when the move elimination optimization didn't
take the fail label of the first select_val into account. This caused it
believe that the only way to reach 'expects_h' was through the second
case when 'Tag' =:= 'h', which made it remove the move instruction
added in the first case, passing garbage to expects_h/2.
|
|
|
|
|
|
|
|
This should be slightly more efficient than converting to/from
lists for large sets.
|
|
|
|
OTP-15633
* siri/rsh-ssh/PR-1787:
Fix some missed comments about rsh
Document the restrictions on the -rsh command
Use ssh as the default remote shell
|
|
|
|
* solvip/stdlib/gen_statem/export-start-types:
Fix links within gen_statem doc
|
|
|
|
Make iolist_size/1 yield
OTP-15631
|
|
|
|
OTP-14831
|
|
* solvip/stdlib/gen_statem/export-start-types:
Create a gen_statem type for enter_loop options
gen_statem exports types related to starting & naming
|
|
|
|
* maint:
mnesia: Avoid raise in overload reports
|
|
* dgud/mnesia/overload-raise/ERIERL-310/OTP-15619:
mnesia: Avoid raise in overload reports
|
|
The iolist_size/1 function did not yield even if the input list was
very long and a call to the function did only consume a single
reduction. This commit fixes these problems.
|
|
|
|
Clarify and avoid misunderstandings. In the previous sentence,
it is written that "the test case is the smallest unit...". To me
it was confusing when I read the following sentence: "Subsets of
test cases, ...". The key part of the paragraph is that test cases
can be grouped, not that they are subsets.
|
|
Add net and socket (modules) to the list of preloaded
modules (returned by the systools_make:preloaded/0
function, why?).
OTP-14831
|
|
OTP-15621 OTP-15622
* rickard/deprecations-removals:
Document deprecations and removals
|
|
|
|
* hans/crypto/new_api_rc1/OTP-15451:
crypto: New experimental api
|
|
The new files api_ng.h and api_ng.c implements an api using EVP.
The api is not by any mean new, except for the crypto application
in Erlang/OTP.
The aims at using the block api in a stream manor, that is
1) call crypto_init/4
2..N) call crypto_update/{2,3}
The purpose is to simplify and hopefully optimize the SSL and SSH
applications.
By keeping the crypto state in C in an enif_resource the costful state
copying in SSL and SSH is reduced with 1-2 per message sent or received.
Changes in other files are for adaptation like FIPS etc since many
functions uses the central get_cipher_type() function.
|
|
* peterdmv/ssl/dtls_logging:
ssl: Add debug logging for DTLS
Change-Id: I83bf117c6c3428c57010e0e581775dd941fc829a
|
|
Fix missing double quote in filename documentation
|
|
* maint:
dialyzer: Fix maps as parameter of opaque types
dialyzer: Fix key check of lists:key{search,member,find}()
|
|
* dialyzer/hasse/fix_bug_erl_bif_types/OTP-15570:
dialyzer: Fix maps as parameter of opaque types
dialyzer: Fix key check of lists:key{search,member,find}()
|
