Bug fix for D154102, where multi-line labels would get incorrect header splitter line in CFG dot

[mark-sed]

Summary

Currently when control flow dot graph is generated for a function with basic block name longer than 80 lines, then the header separator will be placed after the line break for the label name instead of after the whole label name. The fix is simple by just moving the insert of | character before the line splitting happens.

A test for this was also added.

[mark-sed]

Fixed dot-scops-npm.ll test.

[apilipenko]

llvm/test/Other/cfg-multiline-header.ll

17

Nit. Please, add a new line.

[mark-sed]

Added new line at the end of the test file

llvm/include/llvm/Analysis/CFGPrinter.h

std::string CompleteNodeLabelString(
   std::string Str;
   raw_string_ostream OS(Str);
   HandleBasicBlock(OS, *Node);
   std::string OutStr = OS.str();
   // Remove "%" from BB name
   if (OutStr[0] == '%') {
     OutStr.erase(OutStr.begin());
   }
+  // Place | after BB name to separate it into header
+  OutStr.insert(OutStr.find_first_of('\n') + 1, "\\|");
 
   unsigned ColNum = 0;
   unsigned LastSpace = 0;
   for (unsigned i = 0; i != OutStr.length(); ++i) {
     if (OutStr[i] == '\n') { // Left justify
       OutStr[i] = '\\';
       OutStr.insert(OutStr.begin() + i + 1, 'l');
       ColNum = 0;
if (OutStr[i] == '\n') { // Left justify
       ColNum = i - LastSpace;
       LastSpace = 0;
       i += 3; // The loop will advance 'i' again.
     } else
       ++ColNum;
     if (OutStr[i] == ' ')
       LastSpace = i;
   }
-  // Replace \l after BB name with | to separate it into header
-  OutStr.replace(OutStr.find_first_of('\\') + 1, 1, "|");
   return OutStr;
 }
 
 template <>
 struct DOTGraphTraits<DOTFuncInfo *> : public DefaultDOTGraphTraits {
 
   // Cache for is hidden property
   DenseMap<const BasicBlock *, bool> isOnDeoptOrUnreachablePath;

llvm/test/Analysis/DotMachineCFG/AMDGPU/functions.mir

 # RUN: llc -mtriple=amdgcn-- -run-pass=dot-machine-cfg  -mcfg-dot-filename-prefix=%t -mcfg-func-name=func2 -o -  %s 2>&1 > /dev/null
 # RUN: FileCheck %s -input-file=%t.func2.dot --check-prefix=MCFG
 
 # MCFG-NOT: digraph "Machine CFG for 'func1' function"
 name: func1
 body: |
   bb.0:
     $sgpr0 = S_LOAD_DWORD_IMM $sgpr10_sgpr11, 0, 0
     $sgpr1 = S_LOAD_DWORD_IMM $sgpr12_sgpr13, 0, 0
     S_ENDPGM 0
 ...
 
 # MCFG: digraph "Machine CFG for 'func2' function"
 # MCFG-NEXT: label="Machine CFG for 'func2' function"
-# MCFG: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.0:| $sgpr0 = S_LOAD_DWORD_IMM $sgpr12_sgpr13, 0, 0\l $sgpr1 = S_LOAD_DWORD_IMM $sgpr6_sgpr7, 0, 0\l $sgpr2 = S_LOAD_DWORD_IMM $sgpr14_sgpr15, 0, 0\l S_ENDPGM 0\l}"];
+# MCFG: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.0:\l| $sgpr0 = S_LOAD_DWORD_IMM $sgpr12_sgpr13, 0, 0\l $sgpr1 = S_LOAD_DWORD_IMM $sgpr6_sgpr7, 0, 0\l $sgpr2 = S_LOAD_DWORD_IMM $sgpr14_sgpr15, 0, 0\l S_ENDPGM 0\l}"];
 ---
 name: func2
 body: |
   bb.0:
     $sgpr0 = S_LOAD_DWORD_IMM $sgpr12_sgpr13, 0, 0
     $sgpr1 = S_LOAD_DWORD_IMM $sgpr6_sgpr7, 0, 0
     $sgpr2 = S_LOAD_DWORD_IMM $sgpr14_sgpr15, 0, 0
     S_ENDPGM 0
 ...

llvm/test/Analysis/DotMachineCFG/AMDGPU/irreducible.mir

 # RUN: llc -mtriple=amdgcn-- -run-pass=dot-machine-cfg  -mcfg-dot-filename-prefix=%t -o -  %s 2>&1 > /dev/null
 # RUN: FileCheck %s -input-file=%t.irreducible.dot --check-prefix=MCFG
 # RUN: llc -mtriple=amdgcn-- -run-pass=dot-machine-cfg  -mcfg-dot-filename-prefix=%t -dot-mcfg-only -o -  %s 2>&1 > /dev/null
 # RUN: FileCheck %s -input-file=%t.irreducible.dot --check-prefix=MCFG-ONLY
 
 # MCFG: digraph "Machine CFG for 'irreducible' function"
 # MCFG-NEXT: label="Machine CFG for 'irreducible' function"
-# MCFG: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.0:| successors: %bb.1(0x40000000), %bb.2(0x40000000)\l  liveins: $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9,\l... $sgpr10_sgpr11, $sgpr14, $sgpr15, $sgpr16\l  %0:sreg_32 = IMPLICIT_DEF\l  %1:vgpr_32 = COPY $vgpr0\l  %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec\l  S_CMP_EQ_U32 %0:sreg_32, 0, implicit-def $scc\l  S_CBRANCH_SCC1 %bb.1, implicit $scc\l  S_BRANCH %bb.2\l}"];
+# MCFG: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.0:\l| successors: %bb.1(0x40000000), %bb.2(0x40000000)\l  liveins: $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9,\l... $sgpr10_sgpr11, $sgpr14, $sgpr15, $sgpr16\l  %0:sreg_32 = IMPLICIT_DEF\l  %1:vgpr_32 = COPY $vgpr0\l  %2:vgpr_32 = V_MOV_B32_e32 0, implicit $exec\l  S_CMP_EQ_U32 %0:sreg_32, 0, implicit-def $scc\l  S_CBRANCH_SCC1 %bb.1, implicit $scc\l  S_BRANCH %bb.2\l}"];
 # MCFG-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
 # MCFG-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
-# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.1:|\l successors: %bb.3(0x80000000)\l\l  %3:vgpr_32 = PHI %2:vgpr_32, %bb.0, %4:vgpr_32, %bb.5\l  %5:vgpr_32 = V_ADD_U32_e64 %3:vgpr_32, 1, 0, implicit $exec\l  S_BRANCH %bb.3\l}"];
+# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.1:\l|\l successors: %bb.3(0x80000000)\l\l  %3:vgpr_32 = PHI %2:vgpr_32, %bb.0, %4:vgpr_32, %bb.5\l  %5:vgpr_32 = V_ADD_U32_e64 %3:vgpr_32, 1, 0, implicit $exec\l  S_BRANCH %bb.3\l}"];
 # MCFG-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
-# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.2:|\l successors: %bb.3(0x80000000)\l\l  %6:vgpr_32 = PHI %2:vgpr_32, %bb.0, %4:vgpr_32, %bb.4\l  %7:vgpr_32 = V_ADD_U32_e64 %6:vgpr_32, 2, 0, implicit $exec\l}"];
+# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.2:\l|\l successors: %bb.3(0x80000000)\l\l  %6:vgpr_32 = PHI %2:vgpr_32, %bb.0, %4:vgpr_32, %bb.4\l  %7:vgpr_32 = V_ADD_U32_e64 %6:vgpr_32, 2, 0, implicit $exec\l}"];
 # MCFG-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
-# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.3:|\l successors: %bb.4(0x80000000)\l\l  %4:vgpr_32 = PHI %5:vgpr_32, %bb.1, %7:vgpr_32, %bb.2\l}"];
+# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.3:\l|\l successors: %bb.4(0x80000000)\l\l  %4:vgpr_32 = PHI %5:vgpr_32, %bb.1, %7:vgpr_32, %bb.2\l}"];
 # MCFG-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
-# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.4:|\l successors: %bb.2(0x40000000), %bb.5(0x40000000)\l\l  %8:vgpr_32 = V_AND_B32_e32 3, %1:vgpr_32, implicit $exec\l  %9:sreg_64 = V_CMP_EQ_U32_e64 %8:vgpr_32, 2, implicit $exec\l  %10:sreg_64 = SI_IF killed %9:sreg_64, %bb.2, implicit-def dead $exec,\l... implicit-def dead $scc, implicit $exec\l}"];
+# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.4:\l|\l successors: %bb.2(0x40000000), %bb.5(0x40000000)\l\l  %8:vgpr_32 = V_AND_B32_e32 3, %1:vgpr_32, implicit $exec\l  %9:sreg_64 = V_CMP_EQ_U32_e64 %8:vgpr_32, 2, implicit $exec\l  %10:sreg_64 = SI_IF killed %9:sreg_64, %bb.2, implicit-def dead $exec,\l... implicit-def dead $scc, implicit $exec\l}"];
 # MCFG-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
 # MCFG-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
-# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.5:|\l successors: %bb.1(0x40000000), %bb.6(0x40000000)\l\l  %11:sreg_64 = V_CMP_EQ_U32_e64 %8:vgpr_32, 1, implicit $exec\l  %12:sreg_64 = SI_IF killed %11:sreg_64, %bb.1, implicit-def dead $exec,\l... implicit-def dead $scc, implicit $exec\l}"];
+# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.5:\l|\l successors: %bb.1(0x40000000), %bb.6(0x40000000)\l\l  %11:sreg_64 = V_CMP_EQ_U32_e64 %8:vgpr_32, 1, implicit $exec\l  %12:sreg_64 = SI_IF killed %11:sreg_64, %bb.1, implicit-def dead $exec,\l... implicit-def dead $scc, implicit $exec\l}"];
 # MCFG-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
 # MCFG-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
-# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.6:|\l\l S_ENDPGM 0\l}"];
+# MCFG-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{bb.6:\l|\l\l S_ENDPGM 0\l}"];
 
 # MCFG-ONLY: digraph "Machine CFG for 'irreducible' function"
 # MCFG-ONLY-NEXT: label="Machine CFG for 'irreducible' function"
 # MCFG-ONLY: Node{{[0-9A-Za-z]*}} [shape=record,label="{%bb.0}"];
 # MCFG-ONLY-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
 # MCFG-ONLY-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};
 # MCFG-ONLY-NEXT: Node{{[0-9A-Za-z]*}} [shape=record,label="{%bb.1}"];
 # MCFG-ONLY-NEXT: Node{{[0-9A-Za-z]*}} -> Node{{[0-9A-Za-z]*}};

llvm/test/Other/cfg-multiline-header.ll

+; RUN: opt < %s -passes=dot-cfg -cfg-dot-filename-prefix=cfg 2>/dev/null > /dev/null
+; RUN: FileCheck %s -input-file=cfg.foo.dot --check-prefix=CHECK
+
+define void @foo(ptr %A, ptr %B) {
+a_very_long_label_that_should_take_over_eight_symbols_and_span_2_lines_in_cfg_dot_graph:
+; CHECK: label="{a_very_long_label_that_should_take_over_eight_symbols_and_span_2_lines_in_cfg_do\l...t_graph:\l|  store i32 1, ptr %A, align 4\l  store i32 2, ptr %B, align 4\l  br label %short_label\l}"
+  store i32 1, ptr %A
+  store i32 2, ptr %B
+  br label %short_label
+short_label:
+; CHECK: label="{short_label:\l|  br label\l... %an_even_longer_multiline_label_that_will_span_muliple_lines_Lorem_ipsum_dolor\l..._sit_amet_consectetur_adipiscing_elit_sed_do_eiusmod_tempor_incididunt_ut_labor\l...e_et_dolore_magna_aliqua\l}"
+  br label %an_even_longer_multiline_label_that_will_span_muliple_lines_Lorem_ipsum_dolor_sit_amet_consectetur_adipiscing_elit_sed_do_eiusmod_tempor_incididunt_ut_labore_et_dolore_magna_aliqua
+an_even_longer_multiline_label_that_will_span_muliple_lines_Lorem_ipsum_dolor_sit_amet_consectetur_adipiscing_elit_sed_do_eiusmod_tempor_incididunt_ut_labore_et_dolore_magna_aliqua:
+; CHECK: label="{an_even_longer_multiline_label_that_will_span_muliple_lines_Lorem_ipsum_dolor_si\l...t_amet_consectetur_adipiscing_elit_sed_do_eiusmod_tempor_incididunt_ut_labore_e\l...t_dolore_magna_aliqua:\l|  ret void\l}"  
+  ret void
+}

llvm/test/Other/cfg_deopt_unreach.ll

 ; RUN: rm -rf %t
 ; RUN: mkdir -p %t
 ; RUN: opt < %s -passes=dot-cfg -cfg-hide-unreachable-paths -cfg-dot-filename-prefix=%t/unreach 2>/dev/null > /dev/null
 ; RUN: FileCheck %s -input-file=%t/unreach.callee.dot -check-prefix=UNREACH
 ; RUN: opt < %s -passes=dot-cfg -cfg-hide-deoptimize-paths -cfg-dot-filename-prefix=%t/deopt 2>/dev/null > /dev/null
 ; RUN: FileCheck %s -input-file=%t/deopt.callee.dot -check-prefix=DEOPT
 ; RUN: opt < %s -passes=dot-cfg -cfg-dot-filename-prefix=%t/no-flags 2>/dev/null > /dev/null
 ; RUN: FileCheck %s -input-file=%t/no-flags.callee.dot -check-prefix=NO-FLAGS
 ; RUN: opt < %s -passes=dot-cfg -cfg-hide-unreachable-paths -cfg-hide-deoptimize-paths -cfg-dot-filename-prefix=%t/both-flags 2>/dev/null > /dev/null
 ; RUN: FileCheck %s -input-file=%t/both-flags.callee.dot -check-prefix=BOTH-FLAGS
 
 declare i8 @llvm.experimental.deoptimize.i8(...)
 
 define i8 @callee(ptr %c) alwaysinline {
   %c0 = load volatile i1, ptr %c
   br i1 %c0, label %lleft, label %lright
-; NO-FLAGS: label="{lleft:|  %v0 = call i8 (...) @llvm.experimental.deoptimize.i8(i32 1) [ \"deopt\"(i32 1)\l... ]\l  ret i8 %v0\l}"
-; DEOPT-NOT: label="{lleft:|  %v0 = call i8 (...) @llvm.experimental.deoptimize.i8(i32 1) [ \"deopt\"(i32 1)\l... ]\l  ret i8 %v0\l}"
-; UNREACH: label="{lleft:|  %v0 = call i8 (...) @llvm.experimental.deoptimize.i8(i32 1) [ \"deopt\"(i32 1)\l... ]\l  ret i8 %v0\l}"
-; BOTH-FLAGS-NOT: label="{lleft:|  %v0 = call i8 (...) @llvm.experimental.deoptimize.i8(i32 1) [ \"deopt\"(i32 1)\l... ]\l  ret i8 %v0\l}"
+; NO-FLAGS: label="{lleft:\l|  %v0 = call i8 (...) @llvm.experimental.deoptimize.i8(i32 1) [ \"deopt\"(i32\l... 1) ]\l  ret i8 %v0\l}"
+; DEOPT-NOT: label="{lleft:\l|  %v0 = call i8 (...) @llvm.experimental.deoptimize.i8(i32 1) [ \"deopt\"(i32\l... 1) ]\l  ret i8 %v0\l}"
+; UNREACH: label="{lleft:\l|  %v0 = call i8 (...) @llvm.experimental.deoptimize.i8(i32 1) [ \"deopt\"(i32\l... 1) ]\l  ret i8 %v0\l}"
+; BOTH-FLAGS-NOT: label="{lleft:\l|  %v0 = call i8 (...) @llvm.experimental.deoptimize.i8(i32 1) [ \"deopt\"(i32\l... 1) ]\l  ret i8 %v0\l}"
 lleft:
   %v0 = call i8(...) @llvm.experimental.deoptimize.i8(i32 1) [ "deopt"(i32 1) ]
   ret i8 %v0
 
-; NO-FLAGS: label="{lright:|  unreachable\l}"
-; DEOPT: label="{lright:|  unreachable\l}"
-; UNREACH-NOT: label="{lright:|  unreachable\l}"
-; BOTH-FLAGS-NOT: label="{lright:|  unreachable\l}"
+; NO-FLAGS: label="{lright:\l|  unreachable\l}"
+; DEOPT: label="{lright:\l|  unreachable\l}"
+; UNREACH-NOT: label="{lright:\l|  unreachable\l}"
+; BOTH-FLAGS-NOT: label="{lright:\l|  unreachable\l}"
 lright:
   unreachable
 }

polly/test/ScopDetect/dot-scops-npm.ll

 ;        function name and printing a dot file there will be a race condition
 ;        when running tests in parallel.
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 
 define void @func_npm(i32 %n, i32 %m, ptr noalias nonnull %A) {
 ; CHECK:      digraph "Scop Graph for 'func_npm' function"
 ; CHECK-NEXT: label="Scop Graph for 'func_npm' function"
-; CHECK:      Node0x[[EntryID:.*]] [shape=record,label="{entry:|  br label %outer.for\l}"];
+; CHECK:      Node0x[[EntryID:.*]] [shape=record,label="{entry:\l|  br label %outer.for\l}"];
 ; CHECK-NEXT: Node0x[[EntryID]] -> Node0x[[OUTER_FOR_ID:.*]];
 ; CHECK-NEXT: Node0x[[OUTER_FOR_ID]] [shape=record,label="{outer.for:
 ; CHECK-NEXT: Node0x[[OUTER_FOR_ID]] -> Node0x[[INNER_FOR_ID:.*]];
 ; CHECK-NEXT: Node0x[[OUTER_FOR_ID]] -> Node0x[[OUTER_EXIT:.*]];
 ; CHECK-NEXT: Node0x[[INNER_FOR_ID]] [shape=record,label="{inner.for:
 ; CHECK-NEXT: Node0x[[INNER_FOR_ID]] -> Node0x[[BABY1_ID:.*]];
 ; CHECK-NEXT: Node0x[[INNER_FOR_ID]] -> Node0x[[INNER_EXIT_ID:.*]];
 ; CHECK-NEXT: Node0x[[BABY1_ID]] [shape=record,label="{body1: