...with a boost::optional fallback for Xcode < 10 (as std::optional is only
available starting with Xcode 10 according to
<https://en.cppreference.com/w/cpp/compiler_support>, and our baseline for iOS
and macOS is still Xcode 9.3 according to README.md). And mechanically rewrite
all code to use o3tl::optional instead of boost::optional.
One immediate benefit is that disabling -Wmaybe-uninitialized for GCC as per
fed7c3deb3 "Slience bogus
-Werror=maybe-uninitialized" should no longer be necessary (and whose check
happened to no longer trigger for GCC 10 trunk, even though that compiler would
still emit bogus -Wmaybe-uninitialized for uses of boost::optional under
--enable-optimized, which made me ponder whether this switch from
boost::optional to std::optional would be a useful thing to do; I keep that
configure.ac check for now, though, and will only remove it in a follow up
commit).
Another longer-term benefit is that the code is now already in good shape for an
eventual switch to std::optional (a switch we would have done anyway once we no
longer need to support Xcode < 10).
Only desktop/qa/desktop_lib/test_desktop_lib.cxx heavily uses
boost::property_tree::ptree::get_child_optional returning boost::optional, so
let it keep using boost::optional for now.
After a number of preceding commits have paved the way for this change, this
commit is completely mechanical, done with
> git ls-files -z | grep -vz -e '^bin/find-unneeded-includes$' -e '^configure.ac$' -e '^desktop/qa/desktop_lib/test_desktop_lib.cxx$' -e '^dictionaries$' -e '^external/' -e '^helpcontent2$' -e '^include/IwyuFilter_include.yaml$' -e '^sc/IwyuFilter_sc.yaml$' -e '^solenv/gdb/boost/optional.py$' -e '^solenv/vs/LibreOffice.natvis$' -e '^translations$' -e '\.svg$' | xargs -0 sed -i -E -e 's|\<boost(/optional)?/optional\.hpp\>|o3tl/optional.hxx|g' -e 's/\<boost(\s*)::(\s*)(make_)?optional\>/o3tl\1::\2\3optional/g' -e 's/\<boost(\s*)::(\s*)none\>/o3tl\1::\2nullopt/g'
(before committing include/o3tl/optional.hxx, and relying on some GNU features).
It excludes some files where mention of boost::optional et al should apparently
not be changed (and the sub-repo directory stubs). It turned out that all uses
of boost::none across the code base were in combination with boost::optional, so
had all to be rewritten as o3tl::nullopt.
Change-Id: Ibfd9f4b3d5a8aee6e6eed310b988c4e5ffd8b11b
Reviewed-on: https://gerrit.libreoffice.org/84128
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
...following up on 4f8a744c4f "Make
loplugin:unnecessaryparen treat member expressions consistently"
Change-Id: I444d2995e88990c3c6fa2b912ef68032daf2cad9
Reviewed-on: https://gerrit.libreoffice.org/84112
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
Stumbled across a warning starting to get emitted for some
!(p->x)
when I temporarily changed x from boost::optional (which has a member operator!)
to std::optional (which instead implicitly uses a member conversion operator to
bool). (That is, for the new
static int foo3(Foo2 & foo) {
(void) !(foo.p);
(void) !(foo.b);
(void) !(foo.n);
return (foo.p) ? 1 : 0;
}
test, the first, third, and fourth body lines never warned, while the second one
erroneously warned without this fix.)
Change-Id: I60f6941aaa3a73db0f1373c954e47aa68d703170
Reviewed-on: https://gerrit.libreoffice.org/84079
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
When building code using Skia, macros for Skia are defined
in config_skia.h, and Skia uses #ifdef to check for them, but that's
ok, and we neither care nor can do much about it.
Change-Id: I0029f0d82b26638e0395efec06c03303333ec1e9
...since bfaf25f593 "Silence
-Werror,-Wunused-variable (clang-cl)" taught it about leading [[maybe_unused]].
Change-Id: I0d3baf66623c50cfb35d3923273fb1129f66213f
Reviewed-on: https://gerrit.libreoffice.org/83728
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
We already use that normally, the build system should generally take
care of this by gmake dependencies. One exception is system headers,
but those usually do not really change even if their timestamps do,
so this should prevent Clang from giving an error about this after
system headers update (which should be harmless, and if it isn't
then I suppose either something else such as compiler upgrade
should trigger a rebuild, or cleaning needs to be done manually).
Change-Id: I60581c0d2d3233aadc60e1c90bd62cb9ad684ba2
Reviewed-on: https://gerrit.libreoffice.org/83662
Tested-by: Jenkins
Reviewed-by: Luboš Luňák <l.lunak@collabora.com>
...following up on 314f15bff0 "Extend
loplugin:external to warn about enums".
Cases where free functions were moved into an unnamed namespace along with a
class, to not break ADL, are in:
filter/source/svg/svgexport.cxx
sc/source/filter/excel/xelink.cxx
sc/source/filter/excel/xilink.cxx
svx/source/sdr/contact/viewobjectcontactofunocontrol.cxx
All other free functions mentioning moved classes appear to be harmless and not
give rise to (silent, even) ADL breakage. (One remaining TODO in
compilerplugins/clang/external.cxx is that derived classes are not covered by
computeAffectedTypes, even though they could also be affected by ADL-breakage---
but don't seem to be in any acutal case across the code base.)
For friend declarations using elaborate type specifiers, like
class C1 {};
class C2 { friend class C1; };
* If C2 (but not C1) is moved into an unnamed namespace, the friend declaration
must be changed to not use an elaborate type specifier (i.e., "friend C1;"; see
C++17 [namespace.memdef]/3: "If the name in a friend declaration is neither
qualified nor a template-id and the declaration is a function or an
elaborated-type-specifier, the lookup to determine whether the entity has been
previously declared shall not consider any scopes outside the innermost
enclosing namespace.")
* If C1 (but not C2) is moved into an unnamed namespace, the friend declaration
must be changed too, see <https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71882>
"elaborated-type-specifier friend not looked up in unnamed namespace".
Apart from that, to keep changes simple and mostly mechanical (which should help
avoid regressions), out-of-line definitions of class members have been left in
the enclosing (named) namespace. But explicit specializations of class
templates had to be moved into the unnamed namespace to appease
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92598> "explicit specialization of
template from unnamed namespace using unqualified-id in enclosing namespace".
Also, accompanying declarations (of e.g. typedefs or static variables) that
could arguably be moved into the unnamed namespace too have been left alone.
And in some cases, mention of affected types in blacklists in other loplugins
needed to be adapted.
And sc/qa/unit/mark_test.cxx uses a hack of including other .cxx, one of which
is sc/source/core/data/segmenttree.cxx where e.g. ScFlatUInt16SegmentsImpl is
not moved into an unnamed namespace (because it is declared in
sc/inc/segmenttree.hxx), but its base ScFlatSegmentsImpl is. GCC warns about
such combinations with enabled-by-default -Wsubobject-linkage, but "The compiler
doesn’t give this warning for types defined in the main .C file, as those are
unlikely to have multiple definitions."
(<https://gcc.gnu.org/onlinedocs/gcc-9.2.0/gcc/Warning-Options.html>) The
warned-about classes also don't have multiple definitions in the given test, so
disable the warning when including the .cxx.
Change-Id: Ib694094c0d8168be68f8fe90dfd0acbb66a3f1e4
Reviewed-on: https://gerrit.libreoffice.org/83239
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
...where it failed with
> In file included from C:/lo-clang/core/extensions/source/activex/SOActiveX.cxx:23:
> C:/lo-clang/core/extensions/source/activex/SOActiveX.h(160,1): error: VarDecl, use "bool" instead of 'BOOL' (aka 'int') [loplugin:fakebool]
> BEGIN_MSG_MAP(CSOActiveX)
> ^~~~~~~~~~~~~~~~~~~~~~~~~
> C:/PROGRA~2/MIB055~1/2019/COMMUN~1/VC/Tools/MSVC/1423~1.281/atlmfc/include\atlwin.h(2692,3): note: expanded from macro 'BEGIN_MSG_MAP'
> BOOL bHandled = TRUE; \
> ^~~~~~~~~~~~~~~~~~~~
and no idea why I did the getImmediateMacroCallerLoc call in
913d34ec6b "Extend loplugin:salbool to
loplugin:fakebool". Removing it doesn't cause any issues at least with my local
Linux and Windows --enable-compiler-plugins builds.
Change-Id: I64bd906a34e724aefe980ce8ff2b675e8fb2ec8a
Reviewed-on: https://gerrit.libreoffice.org/83212
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
And while at it remove the unneeded getPlatformGlyphCache
abstraction.
Change-Id: Id5cad751eda9e6bf177dfb4816280d7c5af7066a
Reviewed-on: https://gerrit.libreoffice.org/83125
Tested-by: Jenkins
Reviewed-by: Jan-Marek Glogowski <glogow@fbihome.de>
Errors were:
/home/chris/repos/libreoffice-latest/compilerplugins/clang/sharedvisitor/analyzer.cxx:80:40: error: reference to 'PointerType' is ambiguous
if (auto const t = type->getAs<PointerType>())
/home/chris/repos/libreoffice-latest/compilerplugins/clang/external.cxx:61:44: error: reference to 'PointerType' is ambiguous
else if (auto const t3 = t1->getAs<PointerType>())
Change-Id: Ia5b7add8f2b3160fa3198ed127785bdd61c74796
Reviewed-on: https://gerrit.libreoffice.org/83030
Tested-by: Jenkins
Reviewed-by: Miklos Vajna <vmiklos@collabora.com>
To mitigate the dangers of silently breaking ADL when moving enums into unnamed
namespaces (see the commit message of 206b5b2661
"New loplugin:external"), note all functions that are affected. (The plan is to
extend loplugin:external further to also warn about classes and class templates,
and the code to identify affected functions already takes that into account, so
some parts of that code are not actually relevant for enums.)
But it appears that none of the functions that are actually affected by the
changes in this commit relied on being found through ADL, so no adaptions were
necessary for them.
(clang::DeclContext::collectAllContexts is non-const, which recursively means
that External's Visit... functions must take non-const Decl*. Which required
compilerplugins/clang/sharedvisitor/analyzer.cxx to be generalized to support
such Visit... functions with non-const Decl* parameters.)
Change-Id: Ia215291402bf850d43defdab3cff4db5b270d1bd
Reviewed-on: https://gerrit.libreoffice.org/83001
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
...ignoring uses of "BOOL" in Objective C for now
Change-Id: I3eb48dc8d2592285f03f75dc7bc1b21e2383644d
Reviewed-on: https://gerrit.libreoffice.org/82803
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
...into an unnamed namespace. (And make it into a "leading comment" that
precedes the code it appeartains to. It had orignally been meant as a "trailing
comment" that follows the code it appeartains to, and had been idented further
to indicate that, but clang-format unhelpfully breaks such indentation of
trailing comments that need to be preceded by a line break due to line length
limitations.)
Change-Id: I6a71a6b8e11ba641acfb542142fe14b4b6445f51
Reviewed-on: https://gerrit.libreoffice.org/82792
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
...which internally uses __builtin_expect and thus caused
> core/include/com/sun/star/uno/Any.hxx:750:13: error: implicit conversion of constant 1 of type 'long' to 'bool'; use 'true' instead [loplugin:consttobool]
> assert(false); // this cannot happen
> ^~~~~~~~~~~~~
> /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.15.sdk/usr/include/assert.h:93:6: note: expanded from macro 'assert'
> (__builtin_expect(!(e), 0) ? __assert_rtn(__func__, __FILE__, __LINE__, #e) : (void)0)
> ^~~~~~~~~~~~~~~~~~~~~~~~~
(See 216bcceee1 "Special handling of
__builtin_expect in boolean expressions" for a similar fix in
loplugin:literaltoboolconversion.)
Change-Id: I4cf4b21006176c908995f5753dd0a38165383d87
Reviewed-on: https://gerrit.libreoffice.org/82773
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
...to: "Find implicit conversions from non-'bool' constants (e.g., 'sal_False')
to 'bool'".
Due to how FALSE is defined as just
#define FALSE (0)
(i.e., a literal of type 'int') but TRUE is defined as
#define TRUE (!FALSE)
(i.e., an implicit conversion from 'int' to 'bool') in GLib (see the comment in
ConstToBool::VisitImplicitCastExpr), we get more warnings about uses of 'TRUE'
than of 'FALSE'. For example, in libreofficekit/source/gtk/lokdocview.cxx there
is a warning about the 'TRUE' in
g_main_context_iteration(nullptr, TRUE);
but not about the 'FALSE' in
g_main_context_iteration(nullptr, FALSE);
(where the parameter of 'g_main_context_iteration' is of type 'gboolean'). Lets
live with that asymmetry for now...
(Besides the issues addressed directly in this commit, it also found the two
bogus asserts at 7e09d08807 "Fix useless
assert(true) (which would never fire)" and
122a0be8ae "Fix useless assert(true) (which would
never fire)", plus 5f0d6df7f5 "Use two-argument
form of static_assert".)
Change-Id: Id77322de9f94b85a7b65608a03e0e9865d14467b
Reviewed-on: https://gerrit.libreoffice.org/82667
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
(When temporarily taking a plugin out of the shared plugin by modifying its
leading
> #ifndef LO_CLANG_SHARED_PLUGINS
line I often forget about a trailing
> #endif // LO_CLANG_SHARED_PLUGINS
line, which then caused a hard to understand
> [GEN] compilerplugins/clang/sharedvisitor/sharedvisitor.cxx
> Incorrect version ''
failure. This way, at least somehow mention the name of the problematic plugin
to give a bit of a clue.)
Change-Id: I9a9178ba0d7a32f448e91076376989f088ea70f8
Reviewed-on: https://gerrit.libreoffice.org/82657
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
...checking for unnecessary uses of more "fake bool" types.
In the past, some of the checks involving the types of variables or data
members, or the return types of functions, issued warnings that required
surrounding code to be changed too (e.g., when changing the signature of a
function whose address was taken). These checks have been tightened now to not
warn in such cases (which avoids warnings that require changes to additional
code, or changes that might even be impossible to make, at the cost of being
less aggressive about removing all unnecessary uses of those "fake bool" types).
Change-Id: I70eb75039817cda34ed611387ee27dc5f36a3e2e
Reviewed-on: https://gerrit.libreoffice.org/82554
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
(I'm planning to use it for a FilteringRewritePlugin.)
And while at it, base its current uses on FilteringPlugin.
Change-Id: I0acdcc6cb0b3a434b425405c8c438dbf65e4d3cb
Reviewed-on: https://gerrit.libreoffice.org/82451
Tested-by: Jenkins
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
See comments at <https://gerrit.libreoffice.org/#/c/81958/> "Only initialize
function pointer once" for a case where such a false warning caused trouble (in
a lambda with deduced return type and multiple return statements).
Change-Id: I64b0b8c45bd3d2a6075e336c361ec778fa0da481
Reviewed-on: https://gerrit.libreoffice.org/82034
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>
