Re: [Cocci] [PATCH] parsing_hacks: Add bool to list of known typedefs
On Tue, 2020-01-21 at 06:23 +, Julia Lawall wrote: > > On Sun, 19 Jan 2020, Jaskaran Singh wrote: > > > bool is widely used in the Linux kernel. Certain cases of > > parsing_hacks.ml would mislabel bool. > > > > Add bool to the list of known typedefs. > > As far as I can see, this function is only used for avoiding to print > some > warning messages. So I'm not sure why it would have any impact on > the > parsing at all? > Sorry, It should've been clear in the patch. This patch will only make any difference if it's applied on top of https://www.mail-archive.com/cocci@systeme.lip6.fr/msg06706.html Cheers, Jaskaran. > julia > > > Stats of --parse-c on Linux v5.5-rc4 are as follows: > > > > Before: > > > > nb good = 18956150, nb passed = 134062 => 0.70% passed > > > > After: > > > > nb good = 18956150, nb passed = 134073 => 0.70% passed > > > > Signed-off-by: Jaskaran Singh > > --- > > The above diff has been obtained after applying > > https://www.mail-archive.com/cocci@systeme.lip6.fr/msg06706.html > > > > parsing_c/parsing_hacks.ml | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/parsing_c/parsing_hacks.ml > > b/parsing_c/parsing_hacks.ml > > index 685a4908..5e3301a0 100644 > > --- a/parsing_c/parsing_hacks.ml > > +++ b/parsing_c/parsing_hacks.ml > > @@ -61,7 +61,7 @@ let is_known_typdef = > >| "u_char" | "u_short" | "u_int" | "u_long" > >| "u8" | "u16" | "u32" | "u64" > >| "s8" | "s16" | "s32" | "s64" > > - | "__u8" | "__u16" | "__u32" | "__u64" > > + | "__u8" | "__u16" | "__u32" | "__u64" | "bool" > > -> true > > > >| "acpi_handle" > > -- > > 2.21.1 > > > > ___ Cocci mailing list Cocci@systeme.lip6.fr https://systeme.lip6.fr/mailman/listinfo/cocci
Re: [Cocci] [PATCH] parsing_hacks: Add bool to list of known typedefs
On Tue, 21 Jan 2020, Jaskaran Singh wrote: > On Tue, 2020-01-21 at 06:20 +, Julia Lawall wrote: > > > > On Sun, 19 Jan 2020, Jaskaran Singh wrote: > > > > > bool is widely used in the Linux kernel. Certain cases of > > > parsing_hacks.ml would mislabel bool. > > > > > > Add bool to the list of known typedefs. > > > > > > Stats of --parse-c on Linux v5.5-rc4 are as follows: > > > > > > Before: > > > > > > nb good = 18956150, nb passed = 134062 => 0.70% passed > > > > > > After: > > > > > > nb good = 18956150, nb passed = 134073 => 0.70% passed > > > > Why did nb passed increase? These are tokens that will not be able > > to be > > transformed by a semantic patch. > > > > I suspect it's instances of nokprobe_inline in > linux/kernel/trace/trace_kprobe.c. They're marked as CppMacro. OK, that looks ok. Could you update the log messages with that information? thanks, julia > > Here's results from my diff: > > 20402,20406c20288,20294 > < passed:bool > < passed:bool > < passed:bool > < passed:bool > < passed:bool > --- > > passed:nokprobe_inline > > passed:nokprobe_inline > > passed:nokprobe_inline > > passed:nokprobe_inline > > passed:nokprobe_inline > > passed:nokprobe_inline > > passed:nokprobe_inline > 20407a20296,20297 > > passed:nokprobe_inline > > passed:nokprobe_inline > 20408a20299,20300 > > passed:nokprobe_inline > > passed:nokprobe_inline > 20409a20302,20303 > > passed:nokprobe_inline > > passed:nokprobe_inline > 20410a20305,20306 > > passed:nokprobe_inline > > passed:nokprobe_inline > > Cheers, > Jaskaran. > > > julia > > > > > Signed-off-by: Jaskaran Singh > > > --- > > > The above diff has been obtained after applying > > > https://www.mail-archive.com/cocci@systeme.lip6.fr/msg06706.html > > > > > > parsing_c/parsing_hacks.ml | 2 +- > > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > > > diff --git a/parsing_c/parsing_hacks.ml > > > b/parsing_c/parsing_hacks.ml > > > index 685a4908..5e3301a0 100644 > > > --- a/parsing_c/parsing_hacks.ml > > > +++ b/parsing_c/parsing_hacks.ml > > > @@ -61,7 +61,7 @@ let is_known_typdef = > > >| "u_char" | "u_short" | "u_int" | "u_long" > > >| "u8" | "u16" | "u32" | "u64" > > >| "s8" | "s16" | "s32" | "s64" > > > - | "__u8" | "__u16" | "__u32" | "__u64" > > > + | "__u8" | "__u16" | "__u32" | "__u64" | "bool" > > > -> true > > > > > >| "acpi_handle" > > > -- > > > 2.21.1 > > > > > > > > ___ Cocci mailing list Cocci@systeme.lip6.fr https://systeme.lip6.fr/mailman/listinfo/cocci
Re: [Cocci] [PATCH] parsing_c: Handle case of macro before typedef
On Sun, 19 Jan 2020, Jaskaran Singh wrote: > For the following case: > > > > A case in parsing_hacks.ml sometimes mislabels as a > typedef ident. > > If typedef is a known typedef (such as u8 or *_t), then label > as a CppMacro. Subsequent cases will then label > correctly. OK, I see that you are now using is_known_typdef in a more interesting way. Could you add a test case that shows what has improved? I would still know to know why the number of passed tokens increases. julia > > Following are results of --parse-c on the Linux Kernel v5.5-rc4: > > Before: > > nb good = 18956150, nb passed = 134061 => 0.70% passed > > After: > > nb good = 18956150, nb passed = 134062 => 0.70% passed > > Signed-off-by: Jaskaran Singh > --- > parsing_c/parsing_hacks.ml | 35 +++ > 1 file changed, 35 insertions(+) > > diff --git a/parsing_c/parsing_hacks.ml b/parsing_c/parsing_hacks.ml > index 8374731b..23d675cf 100644 > --- a/parsing_c/parsing_hacks.ml > +++ b/parsing_c/parsing_hacks.ml > @@ -2184,6 +2184,41 @@ let lookahead2 ~pass next before = >&& ok_typedef s && is_macro s2 && is_type type_ > -> > TIdent (s, i1) > + > + (* xx yy zz : xx is a macro *) > + | (TIdent (s, i1)::TIdent (s2, i2)::TIdent(_,_)::_ , _) > +when not_struct_enum before > + && ok_typedef s2 > + && is_known_typdef s2 > +-> > + TCommentCpp(Token_c.CppMacro, i1) > + > + (* xx yy zz : xx is a typedef ident *) > + | (TIdent (s, i1)::TIdent (s2, i2)::TIdent(_,_)::_ , _) > +when not_struct_enum before > + && ok_typedef s > +-> > + msg_typedef s i1 2; LP.add_typedef_root s i1; > + TypedefIdent (s, i1) > + > + (* xx yy * zz : xx is a macro *) > + | (TIdent (s, i1)::TIdent (s2, i2)::ptr , _) > +when pointer ~followed_by:(function TIdent _ -> true | _ -> false) ptr > + && not_struct_enum before > + && ok_typedef s2 > + && is_known_typdef s2 > +-> > + TCommentCpp(Token_c.CppMacro, i1) > + > + (* xx yy * zz : xx is a typedef ident *) > + | (TIdent (s, i1)::TIdent (s2, i2)::ptr , _) > +when pointer ~followed_by:(function TIdent _ -> true | _ -> false) ptr > + && not_struct_enum before > + && ok_typedef s > +-> > + msg_typedef s i1 2; LP.add_typedef_root s i1; > + TypedefIdent (s, i1) > + >(* xx yy *) >| (TIdent (s, i1)::TIdent (s2, i2)::rest , _) when not_struct_enum before > && ok_typedef s && not (is_macro_paren s2 rest) > -- > 2.21.1 > > ___ Cocci mailing list Cocci@systeme.lip6.fr https://systeme.lip6.fr/mailman/listinfo/cocci
Re: [Cocci] [PATCH] parsing_hacks: Add bool to list of known typedefs
On Tue, 2020-01-21 at 06:20 +, Julia Lawall wrote: > > On Sun, 19 Jan 2020, Jaskaran Singh wrote: > > > bool is widely used in the Linux kernel. Certain cases of > > parsing_hacks.ml would mislabel bool. > > > > Add bool to the list of known typedefs. > > > > Stats of --parse-c on Linux v5.5-rc4 are as follows: > > > > Before: > > > > nb good = 18956150, nb passed = 134062 => 0.70% passed > > > > After: > > > > nb good = 18956150, nb passed = 134073 => 0.70% passed > > Why did nb passed increase? These are tokens that will not be able > to be > transformed by a semantic patch. > I suspect it's instances of nokprobe_inline in linux/kernel/trace/trace_kprobe.c. They're marked as CppMacro. Here's results from my diff: 20402,20406c20288,20294 < passed:bool < passed:bool < passed:bool < passed:bool < passed:bool --- > passed:nokprobe_inline > passed:nokprobe_inline > passed:nokprobe_inline > passed:nokprobe_inline > passed:nokprobe_inline > passed:nokprobe_inline > passed:nokprobe_inline 20407a20296,20297 > passed:nokprobe_inline > passed:nokprobe_inline 20408a20299,20300 > passed:nokprobe_inline > passed:nokprobe_inline 20409a20302,20303 > passed:nokprobe_inline > passed:nokprobe_inline 20410a20305,20306 > passed:nokprobe_inline > passed:nokprobe_inline Cheers, Jaskaran. > julia > > > Signed-off-by: Jaskaran Singh > > --- > > The above diff has been obtained after applying > > https://www.mail-archive.com/cocci@systeme.lip6.fr/msg06706.html > > > > parsing_c/parsing_hacks.ml | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/parsing_c/parsing_hacks.ml > > b/parsing_c/parsing_hacks.ml > > index 685a4908..5e3301a0 100644 > > --- a/parsing_c/parsing_hacks.ml > > +++ b/parsing_c/parsing_hacks.ml > > @@ -61,7 +61,7 @@ let is_known_typdef = > >| "u_char" | "u_short" | "u_int" | "u_long" > >| "u8" | "u16" | "u32" | "u64" > >| "s8" | "s16" | "s32" | "s64" > > - | "__u8" | "__u16" | "__u32" | "__u64" > > + | "__u8" | "__u16" | "__u32" | "__u64" | "bool" > > -> true > > > >| "acpi_handle" > > -- > > 2.21.1 > > > > ___ Cocci mailing list Cocci@systeme.lip6.fr https://systeme.lip6.fr/mailman/listinfo/cocci
Re: [Cocci] [PATCH] parsing_hacks: Add bool to list of known typedefs
On Sun, 19 Jan 2020, Jaskaran Singh wrote: > bool is widely used in the Linux kernel. Certain cases of > parsing_hacks.ml would mislabel bool. > > Add bool to the list of known typedefs. As far as I can see, this function is only used for avoiding to print some warning messages. So I'm not sure why it would have any impact on the parsing at all? julia > Stats of --parse-c on Linux v5.5-rc4 are as follows: > > Before: > > nb good = 18956150, nb passed = 134062 => 0.70% passed > > After: > > nb good = 18956150, nb passed = 134073 => 0.70% passed > > Signed-off-by: Jaskaran Singh > --- > The above diff has been obtained after applying > https://www.mail-archive.com/cocci@systeme.lip6.fr/msg06706.html > > parsing_c/parsing_hacks.ml | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/parsing_c/parsing_hacks.ml b/parsing_c/parsing_hacks.ml > index 685a4908..5e3301a0 100644 > --- a/parsing_c/parsing_hacks.ml > +++ b/parsing_c/parsing_hacks.ml > @@ -61,7 +61,7 @@ let is_known_typdef = >| "u_char" | "u_short" | "u_int" | "u_long" >| "u8" | "u16" | "u32" | "u64" >| "s8" | "s16" | "s32" | "s64" > - | "__u8" | "__u16" | "__u32" | "__u64" > + | "__u8" | "__u16" | "__u32" | "__u64" | "bool" > -> true > >| "acpi_handle" > -- > 2.21.1 > > ___ Cocci mailing list Cocci@systeme.lip6.fr https://systeme.lip6.fr/mailman/listinfo/cocci
Re: [Cocci] [PATCH] parsing_hacks: Add bool to list of known typedefs
On Sun, 19 Jan 2020, Jaskaran Singh wrote: > bool is widely used in the Linux kernel. Certain cases of > parsing_hacks.ml would mislabel bool. > > Add bool to the list of known typedefs. > > Stats of --parse-c on Linux v5.5-rc4 are as follows: > > Before: > > nb good = 18956150, nb passed = 134062 => 0.70% passed > > After: > > nb good = 18956150, nb passed = 134073 => 0.70% passed Why did nb passed increase? These are tokens that will not be able to be transformed by a semantic patch. julia > > Signed-off-by: Jaskaran Singh > --- > The above diff has been obtained after applying > https://www.mail-archive.com/cocci@systeme.lip6.fr/msg06706.html > > parsing_c/parsing_hacks.ml | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/parsing_c/parsing_hacks.ml b/parsing_c/parsing_hacks.ml > index 685a4908..5e3301a0 100644 > --- a/parsing_c/parsing_hacks.ml > +++ b/parsing_c/parsing_hacks.ml > @@ -61,7 +61,7 @@ let is_known_typdef = >| "u_char" | "u_short" | "u_int" | "u_long" >| "u8" | "u16" | "u32" | "u64" >| "s8" | "s16" | "s32" | "s64" > - | "__u8" | "__u16" | "__u32" | "__u64" > + | "__u8" | "__u16" | "__u32" | "__u64" | "bool" > -> true > >| "acpi_handle" > -- > 2.21.1 > > ___ Cocci mailing list Cocci@systeme.lip6.fr https://systeme.lip6.fr/mailman/listinfo/cocci
