[llvm] r233207 - Linker: Rewrite dwarfdump checks from r233164
Duncan P. N. Exon Smith
dexonsmith at apple.com
Wed Mar 25 14:41:44 PDT 2015
> On 2015-Mar-25, at 14:20, David Blaikie <dblaikie at gmail.com> wrote:
>
>
>
> On Wed, Mar 25, 2015 at 2:14 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
>
> > On 2015-Mar-25, at 14:11, David Blaikie <dblaikie at gmail.com> wrote:
> >
> >
> >
> > On Wed, Mar 25, 2015 at 2:08 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
> >
> > > On 2015-Mar-25, at 13:06, David Blaikie <dblaikie at gmail.com> wrote:
> > >
> > >
> > >
> > > On Wed, Mar 25, 2015 at 12:57 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
> > > Author: dexonsmith
> > > Date: Wed Mar 25 14:57:42 2015
> > > New Revision: 233207
> > >
> > > URL: http://llvm.org/viewvc/llvm-project?rev=233207&view=rev
> > > Log:
> > > Linker: Rewrite dwarfdump checks from r233164
> > >
> > > Rewrite the checks from r233164 that I temporarily disabled in r233165.
> > >
> > > It turns out that the line-tables only debug info we emit from `llc` is
> > > (intentionally) different on Linux than on Darwin.
> > >
> > > Does this need to test gmlt rather than normal -g?
> >
> > I suppose not, but it's easier to hand-write line-tables only debug info,
>
> Really?! I had assumed the backend required more fidelity or details or
> something when not in `-gmlt` mode. I'll play around.
>
> Nah - in fact there was no backend option for gmlt until Eric added it to avoid emitting aranges, I think? Prior to that it was purely a frontend feature that just skipped creating the metadata for types, parameters, variables, etc.
>
> Then I used the flag Eric added to make the gmlt even smarter in the backend (by omitting unused subprograms, omitting unnecessary attributes, etc).
Awesome. Hopefully finally finished with this blasted testcase with
r233221, which tests the debug info directly rather than relying on
links to the line table. I left in checks of the line table itself
since we care about that as well, but they're no longer codependent.
>
>
> >
> > You can create minimal debug inof by hand without making it "gmlt" debug info - don't set that gmlt flag on the CU and it shouldn't do the magical "minimal subprogram" thing on non-darwin anymore.
> >
> > and that should be sufficient for checking if the correct subprogram is
> > around.
> >
> > >
> > > If it does need to test gmlt, which bits of it does it need to test? Is the functionality you want to test actually used in both Darwin and non-Darwin, or is it only under the non-Darwin "skip unnecessary subprogram" stuff I implemented? (in which cas running the test on Darwin doesn't actually tell you anything about whether the functionality is working - so it might be more suitable to use a fixed triple?)
> >
> > I want to test:
> >
> > 1. That the merged IR zeros out Functions (although I imagine this could
> > change eventually).
> > 2. That `llc` doesn't crash.
> > 3. That in the output DWARF, the debug info from the `weak` version of
> > `@foo` is what's actually used (not the debug info from the discarded
> > `linkonce` version of `@foo`).
> >
> > All these things should be true for both Linux and Darwin (and any other
> > platform we support, I think).
> >
> > >
> > > r218129 started
> > > skipping emission of subprograms with no inlined subroutines, and
> > > r218702 was a spiritual revert of that behaviour for Darwin.
> > >
> > > I think we can still test this in a platform-neutral way.
> > >
> > > - Stop checking for the possibly missing `DW_TAG_subprogram` defining
> > > the debug info for the real version of `@foo`.
> > > - Start checking the line tables, ensuring that the right debug info
> > > was used to generate them (grabbing `DW_AT_low_pc` from the compile
> > > unit).
> > > - I changed up the line numbers used in the "weak" version so it's
> > > easier to follow.
> > >
> > > This should hopefully finish off PR22792.
> > >
> > > Modified:
> > > llvm/trunk/test/Linker/Inputs/subprogram-linkonce-weak.ll
> > > llvm/trunk/test/Linker/subprogram-linkonce-weak.ll
> > >
> > > Modified: llvm/trunk/test/Linker/Inputs/subprogram-linkonce-weak.ll
> > > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Linker/Inputs/subprogram-linkonce-weak.ll?rev=233207&r1=233206&r2=233207&view=diff
> > > ==============================================================================
> > > --- llvm/trunk/test/Linker/Inputs/subprogram-linkonce-weak.ll (original)
> > > +++ llvm/trunk/test/Linker/Inputs/subprogram-linkonce-weak.ll Wed Mar 25 14:57:42 2015
> > > @@ -1,7 +1,7 @@
> > > define weak i32 @foo(i32 %a, i32 %b) {
> > > entry:
> > > - %sum = call i32 @fastadd(i32 %a, i32 %b), !dbg !MDLocation(line: 2, scope: !3)
> > > - ret i32 %sum, !dbg !MDLocation(line: 3, scope: !3)
> > > + %sum = call i32 @fastadd(i32 %a, i32 %b), !dbg !MDLocation(line: 52, scope: !3)
> > > + ret i32 %sum, !dbg !MDLocation(line: 53, scope: !3)
> > > }
> > >
> > > declare i32 @fastadd(i32, i32)
> > > @@ -12,5 +12,5 @@ declare i32 @fastadd(i32, i32)
> > > !llvm.dbg.cu = !{!1}
> > > !1 = !MDCompileUnit(language: DW_LANG_C99, file: !2, subprograms: !{!3}, emissionKind: 2)
> > > !2 = !MDFile(filename: "foo.c", directory: "/path/to/dir")
> > > -!3 = !MDSubprogram(file: !2, scope: !2, line: 1, name: "foo", function: i32 (i32, i32)* @foo, type: !4)
> > > +!3 = !MDSubprogram(file: !2, scope: !2, line: 51, name: "foo", function: i32 (i32, i32)* @foo, type: !4)
> > > !4 = !MDSubroutineType(types: !{})
> > >
> > > Modified: llvm/trunk/test/Linker/subprogram-linkonce-weak.ll
> > > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Linker/subprogram-linkonce-weak.ll?rev=233207&r1=233206&r2=233207&view=diff
> > > ==============================================================================
> > > --- llvm/trunk/test/Linker/subprogram-linkonce-weak.ll (original)
> > > +++ llvm/trunk/test/Linker/subprogram-linkonce-weak.ll Wed Mar 25 14:57:42 2015
> > > @@ -66,8 +66,8 @@ entry:
> > > ; LW: ![[FOOINBAR]] = !MDLocation(line: 2, scope: ![[FOOSP]], inlinedAt: ![[BARIA:[0-9]+]])
> > > ; LW: ![[BARIA]] = !MDLocation(line: 12, scope: ![[BARSP]])
> > > ; LW: ![[BARRET]] = !MDLocation(line: 13, scope: ![[BARSP]])
> > > -; LW: ![[FOOCALL]] = !MDLocation(line: 2, scope: ![[WEAKFOOSP]])
> > > -; LW: ![[FOORET]] = !MDLocation(line: 3, scope: ![[WEAKFOOSP]])
> > > +; LW: ![[FOOCALL]] = !MDLocation(line: 52, scope: ![[WEAKFOOSP]])
> > > +; LW: ![[FOORET]] = !MDLocation(line: 53, scope: ![[WEAKFOOSP]])
> > >
> > > ; Same as above, but reordered.
> > > ; WL: ![[WCU]] = !MDCompileUnit({{.*}} subprograms: ![[WSPs:[0-9]+]]
> > > @@ -81,8 +81,8 @@ entry:
> > > ; WL: ![[FOOSP]] = {{.*}}!MDSubprogram(name: "foo",
> > > ; Note, for symmetry, this should be "NOT: function:" and "SAME: ){{$}}".
> > > ; WL-SAME: function: i32 (i32, i32)* @foo
> > > -; WL: ![[FOOCALL]] = !MDLocation(line: 2, scope: ![[WEAKFOOSP]])
> > > -; WL: ![[FOORET]] = !MDLocation(line: 3, scope: ![[WEAKFOOSP]])
> > > +; WL: ![[FOOCALL]] = !MDLocation(line: 52, scope: ![[WEAKFOOSP]])
> > > +; WL: ![[FOORET]] = !MDLocation(line: 53, scope: ![[WEAKFOOSP]])
> > > ; WL: ![[FOOINBAR]] = !MDLocation(line: 2, scope: ![[FOOSP]], inlinedAt: ![[BARIA:[0-9]+]])
> > > ; WL: ![[BARIA]] = !MDLocation(line: 12, scope: ![[BARSP]])
> > > ; WL: ![[BARRET]] = !MDLocation(line: 13, scope: ![[BARSP]])
> > > @@ -96,15 +96,18 @@ entry:
> > > ; Crasher for llc.
> > > ; REQUIRES: object-emission
> > > ; RUN: %llc_dwarf -filetype=obj -O0 %t1 -o %t1.o
> > > -; RUNDISABLED: llvm-dwarfdump %t1.o -debug-dump=info | FileCheck %s -check-prefix=DWLW
> > > +; RUN: llvm-dwarfdump %t1.o -debug-dump=all | FileCheck %s -check-prefix=DWLW -check-prefix=DW
> > > ; RUN: %llc_dwarf -filetype=obj -O0 %t2 -o %t2.o
> > > -; RUNDISABLED: llvm-dwarfdump %t2.o -debug-dump=info | FileCheck %s -check-prefix=DWWL
> > > -; Getting different dwarfdump output on different platforms, so I've
> > > -; temporarily disabled the Dwarf FileChecks while leaving in the crash tests.
> > > -; I'll keep using PR22792 to track this.
> > > +; RUN: llvm-dwarfdump %t2.o -debug-dump=all | FileCheck %s -check-prefix=DWWL -check-prefix=DW
> > > +; Don't check for the DW_TAG_subprogram for the weak version of @foo. It has
> > > +; no inlined subroutines, so with -gmlt may not exist (debuggers can find the
> > > +; aranges in the compile unit and look in the line table directly). See the
> > > +; checks against the line table that follow.
> > >
> > > +; DW-LABEL: .debug_info contents:
> > > ; DWLW: DW_TAG_compile_unit
> > > ; DWLW: DW_AT_name {{.*}}"bar.c"
> > > +; DWLW: DW_AT_low_pc [DW_FORM_addr] ([[BARLOW:0x[0-9a-f]+]])
> > > ; DWLW: DW_TAG_subprogram
> > > ; DWLW-NOT: DW_AT_{{[lowhigh]*}}_pc
> > > ; DWLW: DW_AT_name {{.*}}foo
> > > @@ -117,20 +120,15 @@ entry:
> > > ; DWLW: DW_AT_abstract_origin
> > > ; DWLW: DW_TAG_compile_unit
> > > ; DWLW: DW_AT_name {{.*}}"foo.c"
> > > -; DWLW: DW_TAG_subprogram
> > > -; DWLW: DW_AT_low_pc
> > > -; DWLW: DW_AT_high_pc
> > > -; DWLW: DW_AT_name {{.*}}foo
> > > +; DWLW: DW_AT_low_pc [DW_FORM_addr] ([[FOOLOW:0x[0-9a-f]+]])
> > >
> > > ; The DWARF output is already symmetric (just reordered).
> > > ; DWWL: DW_TAG_compile_unit
> > > ; DWWL: DW_AT_name {{.*}}"foo.c"
> > > -; DWWL: DW_TAG_subprogram
> > > -; DWWL: DW_AT_low_pc
> > > -; DWWL: DW_AT_high_pc
> > > -; DWWL: DW_AT_name {{.*}}foo
> > > +; DWWL: DW_AT_low_pc [DW_FORM_addr] ([[FOOLOW:0x[0-9a-f]+]])
> > > ; DWWL: DW_TAG_compile_unit
> > > ; DWWL: DW_AT_name {{.*}}"bar.c"
> > > +; DWWL: DW_AT_low_pc [DW_FORM_addr] ([[BARLOW:0x[0-9a-f]+]])
> > > ; DWWL: DW_TAG_subprogram
> > > ; DWWL-NOT: DW_AT_{{[lowhigh]*}}_pc
> > > ; DWWL: DW_AT_name {{.*}}foo
> > > @@ -141,3 +139,20 @@ entry:
> > > ; DWWL: DW_AT_name {{.*}}bar
> > > ; DWWL: DW_TAG_inlined_subroutine
> > > ; DWWL: DW_AT_abstract_origin
> > > +
> > > +; DW-LABEL: .debug_line contents:
> > > +; Check that the low_pc entries from above hit the right files.
> > > +
> > > +; DWLW-LABEL: file_names[{{ *}}1]{{.*}} bar.c
> > > +; DWLW: [[BARLOW]] 0 0 1 0 0 is_stmt
> > > +; DWLW: [[BARLOW]] 2 0 1 0 0 is_stmt prologue_end
> > > +; DWLW-LABEL: file_names[{{ *}}1]{{.*}} foo.c
> > > +; DWLW: [[FOOLOW]] 0 0 1 0 0 is_stmt
> > > +; DWLW: [[FOOLOW]] 52 0 1 0 0 is_stmt prologue_end
> > > +
> > > +; DWWL-LABEL: file_names[{{ *}}1]{{.*}} foo.c
> > > +; DWWL: [[FOOLOW]] 0 0 1 0 0 is_stmt
> > > +; DWWL: [[FOOLOW]] 52 0 1 0 0 is_stmt prologue_end
> > > +; DWWL-LABEL: file_names[{{ *}}1]{{.*}} bar.c
> > > +; DWWL: [[BARLOW]] 0 0 1 0 0 is_stmt
> > > +; DWWL: [[BARLOW]] 2 0 1 0 0 is_stmt prologue_end
> > >
> > >
> > > _______________________________________________
> > > llvm-commits mailing list
> > > llvm-commits at cs.uiuc.edu
> > > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> > >
More information about the llvm-commits
mailing list