hardened-c++ adds '-fstack-protector','--param=ssp-buffer-size=4' to arguments for compiler only if no -nostdlib or -ffreestanding was passed to compiler.
TorBrowser/Tor/libgmpxx.so
TorBrowser/Tor/libgmpxx.so.4
TorBrowser/Tor/libgmpxx.so.4.3.3
This case is about '-nostdlib' that passed to g++ while libgmpxx compiled. Then hardened-c++ skips ssp arguments.
Si vous recherchez à vous protéger efficacement contre les cambrioleurs, il convient d'investir dans un coffre-fort agréé. Une armoire forte pour armes longues servira éventuellement pour le stockage des fusils de chasse pour ranger vos archives et dossiers importants. Une alarme n'est pas suffisante, les cambrioleurs savent où trouver vos valeurs et liquidités et fond des raids éclairs pour repartir avec vos valeurs, avant que le voisinage ait eu le temps d'alerter la police ou la gendarmerie. Un coffre-fort agréé et homologué qui sera correctement fixé au sol permettra de mettre les cambrioleurs en échec, ils n'auront d'autre choix que de s'enfuir pour ne pas se faire prendre.
Was any of those reported as protected for any previous versions?
hardening-wrapper (1.25) packaged for lucid using -fstack-protector which can't cover any functions from those libs (it needs proof, but brief reading code show that functions are small enough to be protected). If no protected functions then no detection code compiled and no canaries support reported.
hardening-wrapper (1.25)
Btw, it doesn't passes reduced ssp-buffer-size value, which is 8 by default, to protect more functions, with compare to --param ssp-buffer-size=4 used for windows builds.
Comparing Windows and Linux builds, mozalloc lib:
For Windows build, protected one function (exported as _ZN7mozilla14VolatileBuffer4InitEjj)
For Linux build, no functions protected.
Trac: Status: new to needs_information Summary: Some stack canaries are still missing on Tor Browser binaries on Linux to Some stack canaries are still missing on Tor Browser binaries
Stack Protected: When an executable was built without any character arrays being allocated on the stack, this check will lead to false alarms (since there is no use of stack_chk_fail, even though it was compiled with the correct options.
Another one since esr68 is liblgpllibs.so. I guess part of this ticket is to figure first out which of the errors are actually false positives and which need to get addressed. We can then file new tickets for the latter and adapt our test tools accordingly.