Don't create empty elements in include path vars

[dg0yt]

• What does your PR fix?

  This PR fixes a reproducible build error for gettext[tools]:x64-osx when libxml2 was already installed. It may fix a number of similar build quirks with vcpkg_configure_make.

• Which triplets are supported/not supported? Have you updated the CI baseline?

  all (tested on osx), no

• Does your PR follow the maintainer guide?

  yes

• If you have added/updated a port: Have you run ./vcpkg x-add-version --all and committed the result?

  -/-

[dg0yt]

Background

I got reproducible failures when I tried to build gettext[tools] on osx with port libxml2 installed. The error pointed to vcpkg's libxml2 where I would have expected to find usage of gettext's embedded libxml2. From running:

gcc -DHAVE_CONFIG_H -I. -I../.././../src/0.21-508c311661/libtextstyle/lib -I..  -I. -I../.././../src/0.21-508c311661/libtextstyle/lib -I.. -I../.././../src/0.21-508c311661/libtextstyle -Iglib -DIN_LIBTEXTSTYLE -DLIBXML_STATIC    -I../.././../src/0.21-508c311661/libtextstyle/lib/libcroco  -DDEPENDS_ON_LIBICONV=1   -fPIC -g -c ../.././../src/0.21-508c311661/libtextstyle/lib/libxml/globals.c

(no reference to vcpkg!) the errors were like:

../.././../src/0.21-508c311661/libtextstyle/lib/libxml/globals.c:367:7: error: illegal initializer (only variables can be initialized)
void *xmlGenericErrorContext = NULL;
      ^
/Users/me/dev/vcpkg/installed/x64-osx/include/libxml/globals.h:360:4: note: expanded from macro 'xmlGenericErrorContext'
(*(__xmlGenericErrorContext()))
   ^
fatal error: too many errors emitted, stopping now [-ferror-limit=]

I looked into vcpkg_configure_cmake and found:

# Used by GCC
set(ENV{C_INCLUDE_PATH} "${_VCPKG_INSTALLED}/include${VCPKG_HOST_PATH_SEPARATOR}${C_INCLUDE_PATH_BACKUP}")

Note that this creates values like:

/Users/me/dev/vcpkg/installed/x64-osx/include:

... including the trailing :, i.e. a separator followed by an empty element. Gcc documentation states that the empty element is treated as the current directory, aka -I.

Now clang detects duplicate . elements:

clang -cc1 version 12.0.5 (clang-1205.0.22.9) default target x86_64-apple-darwin20.4.0
ignoring nonexistent directory "/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/local/include"
ignoring nonexistent directory "/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/Library/Frameworks"
ignoring duplicate directory "."
ignoring duplicate directory "../.././../src/0.21-508c311661/libtextstyle/lib"
ignoring duplicate directory ".."
ignoring duplicate directory "."
  as it is a non-system directory that duplicates a system directory
#include "..." search starts here:
#include <...> search starts here:
 ../.././../src/0.21-508c311661/libtextstyle/lib
 ..
 ../.././../src/0.21-508c311661/libtextstyle
 glib
 ../.././../src/0.21-508c311661/libtextstyle/lib/libcroco
 /usr/local/include
 /Users/kpa/dev/vcpkg/installed/x64-osx/include
 .
 /Library/Developer/CommandLineTools/usr/lib/clang/12.0.5/include
 /Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/include
 /Library/Developer/CommandLineTools/usr/include
 /Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/System/Library/Frameworks (framework directory)
End of search list.

... and leaves it only after all explicit entries and the vcpkg include dir.

So that is why libxml/globals.h is not taken from . but from /Users/me/dev/vcpkg/installed/x64-osx/include with libxml2 installed.
And that's probably also the reason for #17892 where the earlier definition is from /usr/local/include.

[autoantwort]

Yes this PR fixes #17892. I have tested it locally. 👍

[dg0yt]

x64_osx libmicrohttpd: unrelated, #17653.
x64_uwp libdatrie: unrelated, download error.

[dg0yt]

More background

clang documentation says that "Empty components in the environment variable are ignored." But this is not true, as can be verified by experimentation and source code inspection.

• Empty components at the beginning or end are treated as "."
  https://github.com/llvm/llvm-project/blob/5f58322368b070b63fe2b2559a54f646cb97e2c4/clang/lib/Driver/ToolChains/CommonArgs.cpp#L187-L221
• Unlike CPATH (-I), the components in the per-language variables are added as system include directories (-c-isystem).
  https://github.com/llvm/llvm-project/blob/6381664580080f015bc0c2ec647853f697cf744a/clang/lib/Driver/ToolChains/Clang.cpp#L1389-L1392
• For system include directories, clang removes the earlier duplicates, i.e. those from command line arguments.
  https://github.com/llvm/llvm-project/blob/d480f968ad8b56d3ee4a6b6df5532d485b0ad01e/clang/lib/Frontend/InitHeaderSearch.cpp#L496-L503

Reproducer:

mkdir /tmp/MARK-I-OPTION
mkdir /tmp/MARK-ENV-VAR
echo | C_INCLUDE_PATH=/tmp/MARK-ENV-VAR clang -E -Wp,-v - -o /dev/null -I. -I/tmp/MARK-I-OPTION 2> c_include_path-no-colon 
echo | C_INCLUDE_PATH=/tmp/MARK-ENV-VAR: clang -E -Wp,-v - -o /dev/null -I. -I/tmp/MARK-I-OPTION 2> c_include_path-with-colon
diff -U20 c_include_path-no-colon c_include_path-with-colon
echo | CPATH=/tmp/MARK-ENV-VAR clang -E -Wp,-v - -o /dev/null -I. -I/tmp/MARK-I-OPTION 2> cpath-no-colon 
echo | CPATH=/tmp/MARK-ENV-VAR: clang -E -Wp,-v - -o /dev/null -I. -I/tmp/MARK-I-OPTION 2> cpath-with-colon 
diff -U20 cpath-no-colon cpath-with-colon

Output for C_INCLUDE_PATH (somewhat older clang on Linux):

--- c_include_path-no-colon	2021-05-18 19:23:22.718867910 +0200
+++ c_include_path-with-colon	2021-05-18 19:59:06.197846524 +0200
@@ -1,12 +1,14 @@
 clang -cc1 version 8.0.0 based upon LLVM 8.0.0 default target x86_64-pc-linux-gnu
 ignoring nonexistent directory "/include"
+ignoring duplicate directory "."
+  as it is a non-system directory that duplicates a system directory
 #include "..." search starts here:
 #include <...> search starts here:
- .
  /tmp/MARK-I-OPTION
  /tmp/MARK-ENV-VAR
+ .
  /usr/local/include
  /usr/lib/llvm-8/lib/clang/8.0.0/include
  /usr/include/x86_64-linux-gnu
  /usr/include
 End of search list.

Output for CPATH:

--- cpath-no-colon	2021-05-18 19:23:53.654933769 +0200
+++ cpath-with-colon	2021-05-18 19:24:00.914949225 +0200
@@ -1,12 +1,13 @@
 clang -cc1 version 8.0.0 based upon LLVM 8.0.0 default target x86_64-pc-linux-gnu
 ignoring nonexistent directory "/include"
+ignoring duplicate directory "."
 #include "..." search starts here:
 #include <...> search starts here:
  .
  /tmp/MARK-I-OPTION
  /tmp/MARK-ENV-VAR
  /usr/local/include
  /usr/lib/llvm-8/lib/clang/8.0.0/include
  /usr/include/x86_64-linux-gnu
  /usr/include

Reported as https://bugs.llvm.org/show_bug.cgi?id=50398

[JackBoosY]

CMake Error at D:/installed/x64-uwp/share/yasm-tool-helper/yasm-tool-helper.cmake:16 (message):
  Cross-targetting and x64 ports requiring yasm require the x86-windows
  yasm-tool to be available.  Please install yasm-tool:x86-windows first.
Call Stack (most recent call first):
  ports/mpg123/portfile.cmake:37 (yasm_tool_helper)
  scripts/ports.cmake:141 (include)


Error: Building package mpg123:x64-uwp failed with: BUILD_FAILED
Elapsed time for package mpg123:x64-uwp: 1.738 s

I think we should wait for #16478.

[dg0yt]

gcc (tested: 7.5) behaves the same.

--- c_include_path-no-colon	2021-05-22 18:59:14.961368312 +0200
+++ c_include_path-with-colon	2021-05-22 18:59:14.965368394 +0200
@@ -1,13 +1,15 @@
 ignoring nonexistent directory "/usr/local/include/x86_64-linux-gnu"
 ignoring nonexistent directory "/usr/lib/gcc/x86_64-linux-gnu/7/../../../../x86_64-linux-gnu/include"
+ignoring duplicate directory "."
+  as it is a non-system directory that duplicates a system directory
 #include "..." search starts here:
 #include <...> search starts here:
- .
  /tmp/MARK-I-OPTION
  /tmp/MARK-ENV-VAR
+ .
  /usr/lib/gcc/x86_64-linux-gnu/7/include
  /usr/local/include
  /usr/lib/gcc/x86_64-linux-gnu/7/include-fixed
  /usr/include/x86_64-linux-gnu
  /usr/include
 End of search list.

[dg0yt]

This changes also fixes building gettext[tools] on Linux.

[JackBoosY]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[JackBoosY]

Ping @Neumann-A for review this PR.

[ras0219-msft]

This LGTM, thanks for the fix!