In Eclipse projects that currently have no definite or potential
resource leaks, treat any such diagnostics as errors in the future.
In `com.ibm.wala.core`, enable warnings about definite or potential
resource leaks. Previously these diagnostics were turned off entirely
in this project. So we actually end up with more warnings now than we
had before, but they are all warnings we should eventually look into.
These are all problems that Eclipse can detect, but that it detects no
instances of right now. Treating these as warnings instead of errors
should help prevent us from slipping backward in the future.
Of course, doing nothing isn't always the right behavior. Sometimes a
previously-unhandled value is truly unexpected and one should fail by
throwing an exception. It may not always be clear whether an
exception or doing nothing is the right choice. For some `switch`
statements affected by this commit, I initially guessed that throwing
an exception was the right default behavior, but was proven wrong when
doing so caused WALA regression test failures. That's strong evidence
that the unmatched values were not really unexpected, but merely
should have been handled by doing nothing as before.
Previously each of these `switch` statements would implicitly do
nothing if an unanticipated `enum` value came along. My impression is
that each of these `switch` statements is supposed to be exhaustive,
such that an unexpected (unhandled) value should never appear. If one
does, we should recognize it and complain loudly.
Of course, sometimes the right behavior for previously-unhandled
values is to do nothing. It may not always be clear whether an
exception or doing nothing is the right choice. For this commit,
WALA's regression tests still pass even with the possibility of
throwing an exception for unexpected values. If we assume that the
test suite is thorough, that tells me that throwing an exception is
the right policy for each `switch` statement that I'm changing here.
Casting to `Foo<Bar>` results in an unchecked-cast warning due to Java
generics type erasure. However, sometimes we don't really need a
`Foo<Bar>`, but could simply use any `Foo<?>`. Casting to the latter
creates no warning.
Each of these required careful consideration of what the original
developer *intended* as distinguished from what the developer's code
actually *does*. I believe I got each one right, and WALA's
regression tests agree. A second opinion by a core WALA developer
would be welcome, though.
Most of these are harmless, and are best fixed simply by removing the
redundant check or assignment. The one in FlowType.compareBlocks,
however, revealed a real problem. This code checks for nullness of
`a` *after* having called a method on `a`. Assuming that `a` can
indeed be `null` here, the check must come first to avoid a
`NullPointerException`.
In several places, I saw code of this form:
if (thing == null)
assert thing != null : ... ;
I honestly don't understand the purpose of that `if` statement. Why
not just have the `assert` statement there directly? I removed the
seemingly irrelevant `if` statements in these cases, but if this is
some intentional pattern, please explain it to me.
In a few places where nullness is statically known but non-obvious,
add assert statements to point out what's going on to help future
developers.
Upgrade future such warnings to errors to keep us moving in a cleaner
direction.
We already have plenty of examples of Serializable classes with this
field, and the vast majority of those fields have generated IDs rather
than "1L". From this I infer that using proper serialVersionUID
fields is considered appropriate WALA coding style.
Also report unused variables as errors in the future, not just
warnings. We've fixed all of these as of right now, so let's keep it
clean in the future too.
One such annotation was unnecessary because the thing it was
suppressing no longer happens. Any future unnecessary warning
suppressions of this kind will now be treated as errors.
The other annotations were unnecessary because the corresponding
warnings have been disabled entirely in the Eclipse projects'
configurations. There seems to be no way to tell Eclipse to treat
these as anything other than "info" diagnostics in the future, so
that's how they will remain.
In general, my approach was to try to eliminate each unused parameter
using Eclipse's "Change Method Signature" refactoring. That did not
always succeed: a parameter may be unused in some base class method,
but then be used in subclass's override of that method. In cases
where refactoring to eliminate a parameter failed, I instead annotated
the parameter with '@SuppressWarnings("unused")' to silence the
warning.
Note: this group of changes creates a significant risk of
incompatibility for third-party WALA code. Some removed parameters
change externally-visible APIs. Furthermore, these changes do not
necessarily lead to Java compilation errors. For example, suppose
third-party code subclasses a WALA class or interface, overrides a
method, but does not annotate that method as @Override. Removing a
parameter means that the third-party method no longer overrides. This
can quietly change code behavior without compile-time errors or
warnings. This is exactly why one should use @Override wherever
possible, but we cannot guarantee that third-party WALA users have
done that.
Unnecessary "throws" declarations tend to cascade. If foo() calls
bar() and bar() falsely declares that it might throw IOException, that
often leads a programmer to declare that foo() might throw IOException
as well. Fixing the bar() throws declaration then reveals that we can
fix the foo() throws declaration too. By the time we reach a fixed
point with cleaning these up, we have removed roughly 320 unnecessary
throws declarations.
In a few cases, this cleanup even lets us remove entire "try
... catch" statements where the only thing being caught was an
exception that we now statically know cannot be thrown. Nice!
In Eclipse project configurations, upgrade any future such shenanigans
from warnings to errors. Now that we've fixed this, we don't want it
coming back again.
There is a potential drawback to this change. Conceivably some public
WALA API entry point might have declared that it could throw some
exception merely to reserve the *option* of throwing that exception in
third-party code that subclasses and overrides the API entry point in
question. I have no idea whether this is a significant concern in
practice, though.
I have *not* upgraded this problem to be treated as an error in the
future. Unfortunately Eclipse uses a single configuration setting for
both unnecessary semicolons and also for empty control-flow statements
like `while (p) ;`. I'm not convinced that it's worth rewriting all
instances of the latter into `while (p) { }`. So this is just going
to stay as a warning for now.
The fix is to add "static" where appropriate, of course. I've also
simplified calls to such methods to reflect the fact that they no
longer need a specific object to call the method on.
In projects that contain test inputs, I've left the non-static
declarations unchanged, and instead downgraded the warning to be
ignored. In all other projects, this warning has been upgraded to an
error.
These were not producing warnings in the Eclipse Oxygen GUI, and also
produced no warnings from Tycho when running Maven tests on my local
machine. However, they did result in errors under Travis-CI. I'm not
sure why this inconsistency exists, but hopefully we have now fixed
these raw-type uses in a way that makes everything happy.
There are two such diagnostics: one for collection methods and one for
equals(). See
<https://www.eclipse.org/eclipse/news/4.7/jdt.php#unlikely-argument-types>
for more information about these two new diagnostics.
For each of these diagnostics, I've set the severity level to
"warning" in projects that have some instances of the suspicious code,
or to "error" in projects that have no instances of the suspicious
code.
Generally, overriding one means you should be overriding the other
too.
Also, configure Eclipse to treat any similar cases as errors, rather
than merely warnings.
The additional diagnostics are ones that were previously being
ignored, but which we seem to have been ignoring by default rather
than as a conscious choice.
For diagnostics of which we currently have *zero* instances, treat
these as errors rather than merely warnings. The intent is to
permanently lock out future regressions of things we've completely
fixed. In the future, whenever we fix the last instance of a given
warning in a given Eclipse project, we should also promote that
diagnostic to an error to keep things clean into the future.
In certain cases, one may want to ignore inter-procedural control
dependence. Consider the following example:
flag = getFlagVal();
if (flag) {
doStuff();
}
If we are ignoring interprocedural control dependence, a forward slice
from the first statement will *not* include statements inside doStuff()
and its transitive callees.
This option is useful in scenarios where the effects of statements
inside control-dependent callees can be accounted for via some cheaper
effect analysis. E.g., if you only care about heap effects of control-
dependent callees, you can compute that using mod-ref analysis,
rather than sucking all the control-dependent callee statements into the
slice.
Also added some more detailed comments, a new unit test, and removed
some trailing whitespace.
In all of these cases, the code used to initialize the unused local
seems nontrivial to me. I think this can all be removed without
erasing any needed side effects, but the code might still become
useful to someone in the future. So I'm not really removing the code
entirely, but merely commenting it out.
This fixes 49 Eclipse code style warnings. I'm not sure why these
were overlooked in my previous sweep of missing-@Override warnings.
Ah well; got 'em this time around.
* Fix warnings about unset javacProjectSettings build entries
Specifically, these are all warnings of the form "The
'javacProjectSettings' build entry should be set when there are project
specific compiler settings".
* Add @Override annotations to all methods that do override
This fixes 287 Eclipse code style warnings.
* Cannot add @Override annotations here, so suppress warnings instead
We should be able to add these @Override annotations in the future,
one Eclipse Mars and earlier are no longer supported. For now,
though, they have to go away in order to be compatible with older
Eclipse releases.
Specifically, these are all warnings of the form "The
'javacProjectSettings' build entry should be set when there are project
specific compiler settings".
* Add missing type parameters to overrides of ModRef.makeModVisitor
* Add missing type parameters to overrides of ModRef.makeRefVisitor
This also required slightly generalizing the generic type signature of
the ModRef.makeRefVisitor return type. Instead of returning
"RefVisitor<T, ExtendedHeapModel>", we now return "RefVisitor<T, ?
extends ExtendedHeapModel>". We need that extra flexibility because
at least one override of this method returns a RefVisitor whose second
generic type parameter is a subclass of ExtendedHeapModel.
* Use IAnalysisCacheView instead of AnalysisCache. (#1)
It seems that some additional types need to be changed due to
d24519e974. This may not be inclusive,
however.
* Increase visibility of several methods and constructors.
Used for creating custom contexts and context selectors.
* Ignore the results directory.
* Some generic type fixes in ModRef.java.
Specifically, we're turning off Eclipse warnings about missing version
constraints on required bundles ("Require-Bundle"), exported
packages ("Export-Package"), and imported packages ("Import-Package").
We're not turning these off absolutely everywhere, though: only in
packages where one or more such warnings were actually being reported.
So if a given package was already providing all version constraints
for, say, package imports, then we've kept that warning on in that
package.
Honestly I don't entirely understand the practical implications of
these warnings. However, there were 355 of them across many WALA
subprojects. I take this as evidence that the WALA developers do not
consider these version constraints to be important. Therefore, we may
as well stop warning about something we have no intention of fixing.
That being said, if we *do* want to fix some or all of these, I
welcome any advice on what those fixes should look like. I am rather
ignorant about all things OSGi.
Removing an unused field sometimes means removing constructor code
that used to initialize that field. Removing that initialization code
sometimes leaves whole constructor arguments unused. Removing those
unused arguments can leave us with unused code to compute those
arguments in constructors' callers, and so on. This commit tries to
clean all of this up, working backward from the unused fields that an
earlier commit already removed. Hopefully I have avoided removing
upstream code that had other important side effects, but it wouldn't
hurt for a WALA expert to review this change carefully.
Manu requested that we use this approach instead of adding
`@SuppressWarnings("unused")` at each affected catch block. That
seems reasonable to me, given the large number of such warnings and
the lack of likely harm from ignoring such caught exceptions.
Fixing these Javadoc comments would require adding packages to various
other packages' build paths. In some of the cases suppressed,
changing build paths in that manner would create circular build
dependencies. In other cases, it would simply add a Javadoc-motivated
dependency that does not exist for the real code, which seems
undesirable. For a few cases, the reference seems to be to types in
code we don't even have here, such as code from "android" or
"org.mozilla" packages.
These arise, for example, when Javadoc documentation on a public class
includes a @link to a private field. I can see how this would be
problematic for closed-source Java code where private items are
invisible to outsiders. However, given that WALA is open source,
nothing is truly non-visible. If the WALA documentation authors
considered non-visible references useful when explaining things,
that's fine with me.
We don't turn this warning off in all projects. Rather, we turn it
off only in projects that were producing at least one such warning.
In other words, if a project was already completely "clean" with
respect to this warning, then we leave this warning enabled for that
project.
These changes turn off Eclipse warnings for Javadoc tags without
descriptions. In some subprojects, we turn these off entirely. In
others, leave on missing-descrption checks for "@return" tags only.
We don't turn this warning off in all projects. Rather, we turn it
off only in projects that were producing at least one such warning.
In other words, if a project was already completely "clean" with
respect to this warning, then we leave this warning enabled for that
project.
Turning off these warnings is a partial declaration of Javadoc
bankruptcy. In an ideal world, we would enable and fix all of these
warnings. However, there are 576 of them. Apparently the WALA team's
implicit coding style says that omitting descriptions is OK. If
there's no intent to systematically add descriptions, then we may as
well turn off these warnings so that we can see other warnings that we
may want to fix.
As it turns out, I should have been using "? extends InstanceKey" rather
than "? super InstanceKey". But really, we can just use "?" here since
HeapGraph itself constrains its own type parameter appropriately.
In the cases addressed here, the caught exception was being "handled"
by throwing some new exception. Instead of discarding the old
exception, pass it to the new exception's constructor to indicate the
original cause of the newly-created exception. This practice, called
"exception chaining", can often be useful in debugging.
Note: some of these methods are decidedly nontrivial. Perhaps they
should not actually be removed? If any should be kept around, please
identify them to me. I'll revise this change to retain those methods
and simply annotate them as needed to suppress Eclipse's warning.
Eclipse warns that the "if" statements' true blocks are dead, and
indeed the conditions being tested here can never be true. It's a
little subtle why that's so, though. Changing them to "assert"
statements removes two warnings about deprecated code, while still
helping human readers understand what invariants must hold here.
All of these involve conditionals that check some static, final debug
flag or debug level. The code will indeed be dead if WALA is built
with those debug facilities turned off. But we still want the code
present in case someone needs to turn some aspect of debugging on for
a while.
The supposedly deprecated function (CGNode.addTarget) really seems to
be intended for internal use, not deprecated per se. We are an
internal user here, so presumably it's OK for us to be using this API
entry point.
Ben Liblit
Raffi Khatchadourian
Julian Dolby
Manu Sridharan
Manu Sridharan