avoids the problems of dangling uno singletons invalidated after the first
dispose and the chain of other singletons that don't expect to need to
re-initialize, etc.
reenable editeng cppunit test
inherit i18npool cppunit test from unotest base
drop LibreOfficeProtector, do "throwable" work in setUp/tearDown not
in ctors/dtors
Naming convention for gbuild methods:
- "add" is used for stuff that is logically a part of the target
(i.e. not registered at the Module, but defined in the target's makefile)
- "use" is used for stuff that is logically a different target
(i.e. it is registered at the Module, has it's own makefile, may be
in a different module than the target)
Plain 0 is ambiguous to convert to either long or const char*,
and just adding an overload next to sal_Int32 would be a dupe if
they actually are the same type, so just go with plain int, which
is the sensible thing to do anyway.
sal/qa/rtl_strings/rtl_old_testostring.cxx contains some German comments still
but the code itself uses test strings in German. When the test strings appear
in the comments, they were left unchanged to maintain consistency.
Mark Wolf
The theory is that
- the operator is suboptimal, because it always uses utf-8 as the encoding,
which might possibly lead to some encoding problem somewhere
but
- the SAL_LOG macros need it anyway, or they are otherwise cumbersome
with OUString
- as people learn to use the macros more, rtl/oustringostreaminserter.hxx
will be included in more and more places, eventually possibly triggering
the above problem anyway
- people probably should not just blindly do ostream << oustring if they
have special encoding requirements
So let's try to simply always have the operator available and see how
it works out.
Seeing 791f276833 I've changed my mind,
some embedded \0 can be actually well hidden:
struct foo { const char txt[3]; };
const foos[] = { { "a" }, { "bb" }};
If somebody wants an embedded \0 in a string literal, they need to
say it explicitly by specifying the size.
Avoid ref/unref pair in makeStringAndClear, hook into the
stringbuffer-like 'String' class to expose it's conversion to
immutable strings, and fixup misc. missing instrumentation.
This needs proper fix. It wasn't noticed because it was created by pipe
before afd5eb1f04 and last command
succeeded, so make continued even if compilation failed.
String literals should be reasonably short, and if memory is so low
that they trigger allocation failure, something else would would
trip over very soon anyway.