New Defects reported by Coverity Scan for LibreOffice
caolanm at redhat.com
Fri Aug 7 11:32:47 UTC 2020
On Fri, 2020-08-07 at 08:17 +0200, Stephan Bergmann wrote:
> The above CIDs 1462308--1462314, 1462316, and 1462318 all appear to
> center on the false assumption that TYPELIB_DANGER_GET
> (include/typelib/typedescription.h) could destroy its
> *ppMacroTypeDescr argument (while it will actually only shave off an
> excess refcount via typelib_typedescription_release).
> Question is whether there is a good way to centrally teach Coverity
> Scan about its false assumption there.
It's probable that we had all these before and marked them manually in
the coverity UI as false positive but loplugin:flatten (or somethine
like that) changed the code enough that coverity doesn't recognize it
as the same code anymore and warns about them again.
I don't know a good way to inline mark these up to avoid the warning,
but they are manually set as ignored again and so shouldn't appear in
the next run.
More information about the LibreOffice