When I load https://bugzilla.mozilla.org/ in trunk Firefox, I see three valgrind warnings (two about conditional jump or move depending on an uninitialized value, and the third about use of an uninitialized value) at the following stack (relevant part included; the point is that uninitialized memory is being passed to printf): ... ==4461== by 0x6452743: PR_snprintf (prprf.c:1169) ==4461== by 0x5DCC441: nsString::AppendInt(int, int) (nsStringObsolete.cpp:1215) ==4461== by 0x15A1B37F: nsPK11Token::refreshTokenInfo() (nsPK11TokenDB.cpp:98) ==4461== by 0x15A1B8BA: nsPK11Token::nsPK11Token(PK11SlotInfoStr*) (nsPK11TokenDB.cpp:66) ==4461== by 0x15A1BBCD: nsPK11TokenDB::GetInternalKeyToken(nsIPK11Token**) (nsPK11TokenDB.cpp:481) ==4461== by 0x5DAD8EC: NS_InvokeByIndex_P (xptcinvoke_x86_64_linux.cpp:208) ... These warnings are fixed by the attached patch. I'm quite surprised the code compiled without this patch; I think it's worth considering using stricter compilation options. This is a regression from the changes in attachment 266692 [details] [diff] [review] on bug 217538.

(In reply to comment #0) > I'm quite surprised the code > compiled without this patch; I think it's worth considering using stricter > compilation options. Er, never mind that. Of course it compiled; it was all #ifdef-ed out.

Comment on attachment 308504 [details] [diff] [review] patch r+ (or even killing the whole ifdef completely). bob

cvs commit pkcs11.c Checking in pkcs11.c; /cvsroot/mozilla/security/nss/lib/softoken/pkcs11.c,v <-- pkcs11.c new revision: 1.149; previous revision: 1.148 done