On 2013-08-21 10:40, Niels Thykier wrote: > On 2013-08-15 20:18, Maximiliano Curia wrote: >> [...] >> >> Thanks for the quick reply. >> >> The actual quote is: >> """ >> I think it is quite reasonable. After all, linking without the >> library in question will succeed, and that will leave undefined weak >> symbols undecorated. >> """ >> Available in: >> http://sourceware.org/ml/binutils/2013-07/msg00001.html >> >> Which is referenced in the closing message in #712081. >> >> Happy hacking, >> > > > Mmm, can we presume that these libraries will instead link (directly) > against the libstdc++ ? > > ~Niels > >
Jakub Wilk reminded me over IRC that we cannot rely on it in the general case. However, checking for libstdc++ does reduce the number of false-positives for many of the libraries I checked. So I applied that as a partial fix for now. In other news, while writing a test for this I noticed that if the library is simple enough then --as-needed causes it to end up with no dependency information at all[1]. This causes a shared-lib-without-dependency-information tag. I am guessing this means we should either reduce the certainty of the tag (perhaps even retire it, eventually). ~Niels [1] As an example, try to add -Wl,--as-needed to the g++ line in the Makefile of the binaries-missing-depends-on-libc test. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org