Age | Commit message (Collapse) | Author |
|
It is now OK to inspect and modify the internals of opaque types within
the scope of the module.
The contracts are used for decorating types with opaqueness when it is
harmless to do so. The opaqueness is propagated by the typesig module
and also by the dataflow module.
A lot of details have been fixed or updated. In particular the modules
erl_types and erl_bif_types have been modified extensively.
The version in vsn.mk has been updated to 2.7. The reason is a
modification of #opaque{} in erl_types.
Dialyzer seems to be about five percent slower than it used to be.
|
|
|
|
Dialyzer was constraining bitstring data used in the construction of other
bitstring values too much. These constraints have now been relaxed.
|
|
|
|
* ks/dialyzer-R16B-final-fixes/OTP-10865:
Native code compilation changes
Code simplification
Simplify a case construct using inequality
Simplify a case construct using a boolean operator
Code simplifications + take out obsolete note
|
|
|
|
|
|
The original implementation of the type signature solver is called 'v1'
and the newly introduced alternative implementation is called 'v2'.
It is possible to run just the one of the solvers (in case there is a
bug in for instance the v2 implementation) or both solvers
("--solver v1 --solver v2"). In the latter case an error is thrown if
the outcome differ.
|
|
An alternative implementation of the solver in dialyzer_typesig has
been introduced. It is faster than the original implementation.
Note: there is code for "loop detection". Where a loop occurs, the
evaluation is stopped and the current solution returned. This
behaviour is consistent with how the original implementation works.
There are a few known cases where the loop detection kicks in. They
are due to bugs which will hopefully be fixed in a near future.
|
|
1. Sometimes the solver forgot that a list had entered the error
state. The bug has been fixed by storing the atom 'error' in
MapDict. An example where the bug occurred is
io_lib_pretty:printable_bin(). The returned spec was weaker than it
should have been, but the fix-point loop hid the bug (in this case).
2. lists:partition() has been substituted for lists:splitwith() in
enumerate_constraints(). This fix together with 3. solves a
problem with long execution times for deeply nested fun:s. An
example which is now much faster is
lib/compiler/test/lc_SUITE:deeply_nested/1
(included as dialyzer/test/small_SUITE_data/src/deep_lc.erl).
3. The calculation of components in enumerate_constraints() has been
simplified and optimized. The important thing here is that _all_ of
the simple constraints have been saturated before entering the
complex part.
4. The pretty printing of constraints has been improved.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The relevant terms were instances of Dialyzer's typesig analysis state
that were stored together with some "dynamic" constraints in a dict. Due
to sharing this was not apparent until we tried to copy the dictionary.
Initial debugging was by Björn Gustavsson.
|
|
|
|
Variable substitution was not generalizing any unknown variables.
|
|
Whenever a variable that could hold one of two or more possible closures was
used in a particular application, the application was assumed to fail if ONE of
the closures would fail in this application. This has been corrected to infer
failing application if ALL possible closures would fail in the particular
application.
|
|
|
|
|
|
Dialyzer does not normally emit warnings for functions that
implement non-terminating server loops. This detection failed
when some of the elements in an SCC terminated normally (being
for example list comprehensions or other generic anonymous
functions that were included in the SCC). This patch fixes that.
|
|
Dialyzer had a minor issue when the arguments of the built-in function
'or' had the fixed value 'false'.
This testcase should return no warnings:
-module(false_false).
-export([false_or/0]).
false_or() ->
false or false.
|
|
Nested list comprehensions are translated into strongly connected
funs but dialyzer ignored this. This meant that self-recursive
analysis was normally executed for each fun and the whole fixpoint
took some time to calculate. This patch adds every fun found in
constraint generation to the SCC that is under analysis and then
solves the SCC as a whole.
Possible issues:
- The returned dict contains more entries than usual. This
triggered a bug in contract checking that is dealt with is
this patch as well (dialyzer_contracts.erl).
- As it's not easy to tell apart real fun SCCs from simple funs it
performance might be hampered in the simple funs case.
While doing these changes, also added and strengthened some specs.
In addition, incorporated a modified patch by Ahmed Omar that fixes
a bug in dialyzer_analysis_callgraph module that kept the files of
only one selected directory in dialyzer's GUI.
|
|
Bumped version number and wrote RELEASE_NOTES.
Included all changes described in release notes.
Some spec-related changes to some files in lib/hipe/cerl.
|
|
* ks/hipe:
dialyzer: Fix system_limit exception in race analysis
syntax_tools: Add types and specs for most exported functions
syntax_tools: Support the --enable-native-libs configure option
syntax_tools: Remove $Id$ annotations
dialyzer: New version for the R13B04 release
hipe: Miscellaneous additions
typer: New version for the R13B04 release
Fix a HiPE compiler bug evaluating an expression that throws system_limit
OTP-8460 ks/hipe
|
|
|
|
|