sys::path: Detect posix paths starting with ~ as absolute

[aprantl]

Summary

Turns out there is more than one way a POSIX path can be absolute. Paths starting with ~, such as ~user, or ~/.git are also absolute paths, or at least not relative paths.

Without this, llvm-dwarfdump will append the working directory to paths starting with ~, resulting in paths like ~user/~user.

[mstorsjo]

Hmm. My main question is whether it should be considered absolute in a windows path... We do expand tildes in the native() function into the actual home directory, so in such a context, it is an absolute path there too.

[labath]

I don't believe we should be treating these paths as absolute at such a deep level. ~ is is a purely a userspace artifact used by shells and programs wishing to provide similar UI convenience. It is not recognized by any system calls, nor by any filesystem library that I am aware of. Normally most programs do not even see it, as it will be expanded by the shell. However, if you convince the shell to not expand it, the programs will happily operate on it, as ~ is a valid filename on most filesystems (not that I recommend using it):

$ touch ~/a.cc
$ touch '~/a.cc'
touch: cannot touch '~/a.cc': No such file or directory
$ clang '~/a.cc'
clang-12: error: no such file or directory: '~/a.cc'
clang-12: error: no input files
$ clang ./~/a.cc
clang-12: error: no such file or directory: './~/a.cc'
clang-12: error: no input files
$ clang ~/a.cc
/usr/bin/x86_64-pc-linux-gnu-ld: /usr/lib/gcc/x86_64-pc-linux-gnu/11.2.0/../../../../lib64/crt1.o: in function `_start':
(.text+0x20): undefined reference to `main'
clang-12: error: linker command failed with exit code 1 (use -v to see invocation)
$ mkdir '~'
$ touch '~/a.cc'
$ ls '~'
a.cc

By adding our own special handling of ~ we would be messing with that mechanism. The case of ./~/a.cc is particularly amusing because then sys::path::remove_leading_dotslash will be turning a relative path into an absolute one.

I agree that there are contexts where we would want to treat ~ specially, but this should be opt-in only and based on the context from which that path is coming from. Maybe DW_AT_comp_dir could be one of those contexts. Strictly speaking, it wouldn't be correct, but we do funny things with dwarf paths anyway. IIRC our dwarf code already has some funky logic to detect absolute windows paths on non-windows hosts. Maybe that could be extended to handle ~ as well?

[dexonsmith]

I agree with @labath; we shouldn't change sys::path APIs to do have implicit knowledge of shell-specific behaviours (or to guess at what they might be).

llvm/test/tools/llvm-dwarfdump/AArch64/tilde.s

4

Seems potentially like a bug in whatever generated the DWARF for this location to appear un-expanded; or, potentially a response to an interesting user request, such as specifying -fdebug-prefix-map="some/path=~user".

102

Maybe this testcase could be stripped down to skip the d56b171ee965eba9ba30f4a479a9f2e1703105cf?

[aprantl]

Thanks for the feedback!

llvm/test/tools/llvm-dwarfdump/AArch64/tilde.s

4

This is literally how this testcase was created, yes.

llvm/lib/Support/Path.cpp

 }
 
 StringRef root_path(StringRef path, Style style) {
   const_iterator b = begin(path, style), pos = b, e = end(path);
   if (b != e) {
     bool has_net =
         b->size() > 2 && is_separator((*b)[0], style) && (*b)[1] == (*b)[0];
     bool has_drive = is_style_windows(style) && b->endswith(":");
+    bool has_tilde = is_style_posix(style) && b->startswith("~");
 
     if (has_net || has_drive) {
       if ((++pos != e) && is_separator((*pos)[0], style)) {
         // {C:/,//net/}, so get the first two components.
         return path.substr(0, b->size() + pos->size());
       }
       // just {C:,//net}, return the first component.
       return *b;
     }
 
     // POSIX style root directory.
-    if (is_separator((*b)[0], style)) {
+    if (has_tilde || is_separator((*b)[0], style)) {
       return *b;
     }
   }
 
   return StringRef();
 }
 
 StringRef root_name(StringRef path, Style style) {

 }
 
 StringRef root_directory(StringRef path, Style style) {
   const_iterator b = begin(path, style), pos = b, e = end(path);
   if (b != e) {
     bool has_net =
         b->size() > 2 && is_separator((*b)[0], style) && (*b)[1] == (*b)[0];
     bool has_drive = is_style_windows(style) && b->endswith(":");
+    bool has_tilde = is_style_posix(style) && b->startswith("~");
 
     if ((has_net || has_drive) &&
         // {C:,//net}, skip to the next component.
         (++pos != e) && is_separator((*pos)[0], style)) {
       return *pos;
     }
 
     // POSIX style root directory.
-    if (!has_net && is_separator((*b)[0], style)) {
+    if (!has_net && (has_tilde || is_separator((*b)[0], style))) {
       return *b;
     }
   }
 
   // No path or no root.
   return StringRef();
 }
 

llvm/test/tools/llvm-dwarfdump/AArch64/tilde.s

+# RUN: llvm-mc %s -filetype obj -triple arm64-apple-darwin -o - \
+# RUN: | llvm-dwarfdump - \
+# RUN: | FileCheck %s
+# CHECK: DW_AT_decl_file ("~user/foo/t.c")
+.section    __TEXT,__text,regular,pure_instructions
+.build_version macos, 12, 0
+.globl    _f                              ; -- Begin function f
+.p2align    2
+_f:                                     ; @f
+Lfunc_begin0:
+.file    1 "~user/foo" "t.c"
+.loc    1 1 0                           ; ~user/foo/t.c:1:0
+.cfi_startproc
+; %bb.0:                                ; %entry
+.loc    1 1 11 prologue_end             ; ~user/foo/t.c:1:11
+ret
+Ltmp0:
+Lfunc_end0:
+.cfi_endproc
+                                    ; -- End function
+.section    __DWARF,__debug_abbrev,regular,debug
+Lsection_abbrev:
+.byte    1                               ; Abbreviation Code
+.byte    17                              ; DW_TAG_compile_unit
+.byte    1                               ; DW_CHILDREN_yes
+.byte    37                              ; DW_AT_producer
+.byte    14                              ; DW_FORM_strp
+.byte    19                              ; DW_AT_language
+.byte    5                               ; DW_FORM_data2
+.byte    3                               ; DW_AT_name
+.byte    14                              ; DW_FORM_strp
+.ascii    "\202|"                         ; DW_AT_LLVM_sysroot
+.byte    14                              ; DW_FORM_strp
+.byte    16                              ; DW_AT_stmt_list
+.byte    23                              ; DW_FORM_sec_offset
+.byte    27                              ; DW_AT_comp_dir
+.byte    14                              ; DW_FORM_strp
+.byte    17                              ; DW_AT_low_pc
+.byte    1                               ; DW_FORM_addr
+.byte    18                              ; DW_AT_high_pc
+.byte    6                               ; DW_FORM_data4
+.byte    0                               ; EOM(1)
+.byte    0                               ; EOM(2)
+.byte    2                               ; Abbreviation Code
+.byte    46                              ; DW_TAG_subprogram
+.byte    0                               ; DW_CHILDREN_no
+.byte    17                              ; DW_AT_low_pc
+.byte    1                               ; DW_FORM_addr
+.byte    18                              ; DW_AT_high_pc
+.byte    6                               ; DW_FORM_data4
+.ascii    "\347\177"                      ; DW_AT_APPLE_omit_frame_ptr
+.byte    25                              ; DW_FORM_flag_present
+.byte    64                              ; DW_AT_frame_base
+.byte    24                              ; DW_FORM_exprloc
+.byte    3                               ; DW_AT_name
+.byte    14                              ; DW_FORM_strp
+.byte    58                              ; DW_AT_decl_file
+.byte    11                              ; DW_FORM_data1
+.byte    59                              ; DW_AT_decl_line
+.byte    11                              ; DW_FORM_data1
+.byte    63                              ; DW_AT_external
+.byte    25                              ; DW_FORM_flag_present
+.byte    0                               ; EOM(1)
+.byte    0                               ; EOM(2)
+.byte    0                               ; EOM(3)
+.section    __DWARF,__debug_info,regular,debug
+Lsection_info:
+Lcu_begin0:
+.set Lset0, Ldebug_info_end0-Ldebug_info_start0 ; Length of Unit
+.long    Lset0
+Ldebug_info_start0:
+.short    4                               ; DWARF version number
+.set Lset1, Lsection_abbrev-Lsection_abbrev ; Offset Into Abbrev. Section
+.long    Lset1
+.byte    8                               ; Address Size (in bytes)
+.byte    1                               ; Abbrev [1] 0xb:0x39 DW_TAG_compile_unit
+.long    0                               ; DW_AT_producer
+.short    12                              ; DW_AT_language
+.long    97                              ; DW_AT_name
+.long    111                             ; DW_AT_LLVM_sysroot
+.set Lset2, Lline_table_start0-Lsection_line ; DW_AT_stmt_list
+.long    Lset2
+.long    113                             ; DW_AT_comp_dir
+.quad    Lfunc_begin0                    ; DW_AT_low_pc
+.set Lset3, Lfunc_end0-Lfunc_begin0     ; DW_AT_high_pc
+.long    Lset3
+.byte    2                               ; Abbrev [2] 0x2e:0x15 DW_TAG_subprogram
+.quad    Lfunc_begin0                    ; DW_AT_low_pc
+.set Lset4, Lfunc_end0-Lfunc_begin0     ; DW_AT_high_pc
+.long    Lset4
+                                    ; DW_AT_APPLE_omit_frame_ptr
+.byte    1                               ; DW_AT_frame_base
+.byte    111
+.long    123                             ; DW_AT_name
+.byte    1                               ; DW_AT_decl_file
+.byte    1                               ; DW_AT_decl_line
+                                    ; DW_AT_external
+.byte    0                               ; End Of Children Mark
+Ldebug_info_end0:
+.section    __DWARF,__debug_str,regular,debug
+Linfo_string:
+.asciz    "clang version 14.0.0 ([e-mail]:llvm/llvm-project d56b171ee965eba9ba30f4a479a9f2e1703105cf)" ; string offset=0
+.asciz    "~user/foo/t.c"                 ; string offset=97
+.asciz    "/"                             ; string offset=111
+.asciz    "~user/foo"                     ; string offset=113
+.asciz    "f"                             ; string offset=123
+.section    __DWARF,__apple_names,regular,debug
+Lnames_begin:
+.long    1212240712                      ; Header Magic
+.short    1                               ; Header Version
+.short    0                               ; Header Hash Function
+.long    1                               ; Header Bucket Count
+.long    1                               ; Header Hash Count
+.long    12                              ; Header Data Length
+.long    0                               ; HeaderData Die Offset Base
+.long    1                               ; HeaderData Atom Count
+.short    1                               ; DW_ATOM_die_offset
+.short    6                               ; DW_FORM_data4
+.long    0                               ; Bucket 0
+.long    177675                          ; Hash in Bucket 0
+.set Lset5, LNames0-Lnames_begin        ; Offset in Bucket 0
+.long    Lset5
+LNames0:
+.long    123                             ; f
+.long    1                               ; Num DIEs
+.long    46
+.long    0
+.section    __DWARF,__apple_objc,regular,debug
+Lobjc_begin:
+.long    1212240712                      ; Header Magic
+.short    1                               ; Header Version
+.short    0                               ; Header Hash Function
+.long    1                               ; Header Bucket Count
+.long    0                               ; Header Hash Count
+.long    12                              ; Header Data Length
+.long    0                               ; HeaderData Die Offset Base
+.long    1                               ; HeaderData Atom Count
+.short    1                               ; DW_ATOM_die_offset
+.short    6                               ; DW_FORM_data4
+.long    -1                              ; Bucket 0
+.section    __DWARF,__apple_namespac,regular,debug
+Lnamespac_begin:
+.long    1212240712                      ; Header Magic
+.short    1                               ; Header Version
+.short    0                               ; Header Hash Function
+.long    1                               ; Header Bucket Count
+.long    0                               ; Header Hash Count
+.long    12                              ; Header Data Length
+.long    0                               ; HeaderData Die Offset Base
+.long    1                               ; HeaderData Atom Count
+.short    1                               ; DW_ATOM_die_offset
+.short    6                               ; DW_FORM_data4
+.long    -1                              ; Bucket 0
+.section    __DWARF,__apple_types,regular,debug
+Ltypes_begin:
+.long    1212240712                      ; Header Magic
+.short    1                               ; Header Version
+.short    0                               ; Header Hash Function
+.long    1                               ; Header Bucket Count
+.long    0                               ; Header Hash Count
+.long    20                              ; Header Data Length
+.long    0                               ; HeaderData Die Offset Base
+.long    3                               ; HeaderData Atom Count
+.short    1                               ; DW_ATOM_die_offset
+.short    6                               ; DW_FORM_data4
+.short    3                               ; DW_ATOM_die_tag
+.short    5                               ; DW_FORM_data2
+.short    4                               ; DW_ATOM_type_flags
+.short    11                              ; DW_FORM_data1
+.long    -1                              ; Bucket 0
+.subsections_via_symbols
+.section    __DWARF,__debug_line,regular,debug
+Lsection_line:
+Lline_table_start0:
+

llvm/unittests/Support/Path.cpp

 }
 
 TEST(get_separator, Works) {
   EXPECT_EQ(path::get_separator(path::Style::posix), "/");
   EXPECT_EQ(path::get_separator(path::Style::windows_backslash), "\\");
   EXPECT_EQ(path::get_separator(path::Style::windows_slash), "/");
 }
 
+TEST(is_absolute, Works) {
+  EXPECT_FALSE(path::is_absolute("./foo", path::Style::posix));
+  EXPECT_FALSE(path::is_absolute("~user", path::Style::windows));
+  EXPECT_FALSE(path::is_absolute("~/home", path::Style::windows));
+  EXPECT_TRUE(path::is_absolute("~user", path::Style::posix));
+  EXPECT_TRUE(path::is_absolute("~/home", path::Style::posix));
+}
+
 TEST(is_absolute_gnu, Works) {
   // Test tuple <Path, ExpectedPosixValue, ExpectedWindowsValue>.
   const std::tuple<StringRef, bool, bool> Paths[] = {
       std::make_tuple("", false, false),
       std::make_tuple("/", true, true),
       std::make_tuple("/foo", true, true),
       std::make_tuple("\\", false, true),
       std::make_tuple("\\foo", false, true),
TEST(Support, Path) {
   }
 }
 
 TEST(Support, PathRoot) {
   ASSERT_EQ(path::root_name("//net/hello", path::Style::posix).str(), "//net");
   ASSERT_EQ(path::root_name("c:/hello", path::Style::posix).str(), "");
   ASSERT_EQ(path::root_name("c:/hello", path::Style::windows).str(), "c:");
   ASSERT_EQ(path::root_name("/hello", path::Style::posix).str(), "");
+  ASSERT_EQ(path::root_name("~/hello", path::Style::posix).str(), "");
+  ASSERT_EQ(path::root_name("~hello", path::Style::posix).str(), "");
 
   ASSERT_EQ(path::root_directory("/goo/hello", path::Style::posix).str(), "/");
   ASSERT_EQ(path::root_directory("c:/hello", path::Style::windows).str(), "/");
   ASSERT_EQ(path::root_directory("d/file.txt", path::Style::posix).str(), "");
   ASSERT_EQ(path::root_directory("d/file.txt", path::Style::windows).str(), "");
+  ASSERT_EQ(path::root_directory("~/hello", path::Style::posix).str(), "~");
+  ASSERT_EQ(path::root_directory("~hello", path::Style::posix).str(), "~hello");
 
   SmallVector<StringRef, 40> paths;
   paths.push_back("");
   paths.push_back(".");
   paths.push_back("..");
   paths.push_back("foo");
   paths.push_back("/");
   paths.push_back("/foo");