Re: [Qemu-devel] [qemu-s390x] [PATCH v2] s390x/tcg: add various alignment check
On 02/15/2018 03:49 AM, Cornelia Huck wrote: On Thu, 15 Feb 2018 10:47:45 +0100 David Hildenbrandwrote: On 14.02.2018 20:04, Richard Henderson wrote: On 02/14/2018 09:31 AM, David Hildenbrand wrote: Let's add proper alignment checks for a handful of instructions that require a SPECIFICATION exception in case alignment is violated. Introduce new wout/in functions. Declare them as "static inline" to avoid warnings about not being used for CONFIG_USER_ONLY (as we are right now only using them for privileged instructions). Annoyingly, clang will still warn for this. Hm, so the only solution is to add nasty idfefs then :( Yup, very annoying indeed, but probably the only way to shut clang up... Does marking the function __attribute__((unused)) shut up clang? -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [qemu-s390x] [PATCH v2] s390x/tcg: add various alignment check
On 15.02.2018 15:13, Eric Blake wrote: > On 02/15/2018 03:49 AM, Cornelia Huck wrote: >> On Thu, 15 Feb 2018 10:47:45 +0100 >> David Hildenbrandwrote: >> >>> On 14.02.2018 20:04, Richard Henderson wrote: On 02/14/2018 09:31 AM, David Hildenbrand wrote: > Let's add proper alignment checks for a handful of instructions that > require a SPECIFICATION exception in case alignment is violated. > > Introduce new wout/in functions. Declare them as "static inline" to avoid > warnings about not being used for CONFIG_USER_ONLY (as we are right > now only using them for privileged instructions). Annoyingly, clang will still warn for this. >>> >>> Hm, so the only solution is to add nasty idfefs then :( >> >> Yup, very annoying indeed, but probably the only way to shut clang up... > > Does marking the function __attribute__((unused)) shut up clang? > Guess so, seems to be used in tcg/tcg.c. But I don't think that's a major improvement, as these warnings can actually point you at bugs. -- Thanks, David / dhildenb
Re: [Qemu-devel] [qemu-s390x] [PATCH v2] s390x/tcg: add various alignment check
On 15.02.2018 10:49, Cornelia Huck wrote: > On Thu, 15 Feb 2018 10:47:45 +0100 > David Hildenbrandwrote: > >> On 14.02.2018 20:04, Richard Henderson wrote: >>> On 02/14/2018 09:31 AM, David Hildenbrand wrote: Let's add proper alignment checks for a handful of instructions that require a SPECIFICATION exception in case alignment is violated. Introduce new wout/in functions. Declare them as "static inline" to avoid warnings about not being used for CONFIG_USER_ONLY (as we are right now only using them for privileged instructions). >>> >>> Annoyingly, clang will still warn for this. >>> >> >> Hm, so the only solution is to add nasty idfefs then :( > > Yup, very annoying indeed, but probably the only way to shut clang up... > Will resend soon! -- Thanks, David / dhildenb
Re: [Qemu-devel] [qemu-s390x] [PATCH v2] s390x/tcg: add various alignment check
On Thu, 15 Feb 2018 10:47:45 +0100 David Hildenbrandwrote: > On 14.02.2018 20:04, Richard Henderson wrote: > > On 02/14/2018 09:31 AM, David Hildenbrand wrote: > >> Let's add proper alignment checks for a handful of instructions that > >> require a SPECIFICATION exception in case alignment is violated. > >> > >> Introduce new wout/in functions. Declare them as "static inline" to avoid > >> warnings about not being used for CONFIG_USER_ONLY (as we are right > >> now only using them for privileged instructions). > > > > Annoyingly, clang will still warn for this. > > > > Hm, so the only solution is to add nasty idfefs then :( Yup, very annoying indeed, but probably the only way to shut clang up...
Re: [Qemu-devel] [qemu-s390x] [PATCH v2] s390x/tcg: add various alignment check
On 14.02.2018 20:04, Richard Henderson wrote: > On 02/14/2018 09:31 AM, David Hildenbrand wrote: >> Let's add proper alignment checks for a handful of instructions that >> require a SPECIFICATION exception in case alignment is violated. >> >> Introduce new wout/in functions. Declare them as "static inline" to avoid >> warnings about not being used for CONFIG_USER_ONLY (as we are right >> now only using them for privileged instructions). > > Annoyingly, clang will still warn for this. > Hm, so the only solution is to add nasty idfefs then :( Thanks! > Otherwise, > Reviewed-by: Richard Henderson> > > r~ > -- Thanks, David / dhildenb