Build fixes for z/OS (#695)

Fixes to enable the code to build with a simple `CC=xlc ./configure --enable-ebcdic --disable-unicode`.

Fixes to the tests, so that `make check` passes on EBCDIC platforms.

Add a CI job to do z/OS testing.

Author: NWilson

.github/workflows/build.yml

@@ -281,6 +281,55 @@ jobs:
             ../maint/RunManifestTest install-dir ../maint/manifest-cmakeinstall-solaris
             ../maint/RunSymbolTest install-dir/lib/ ../maint/
 
+  zos:
+    name: z/OS
+    runs-on: ubuntu-latest
+    if: github.event_name != 'pull_request'
+    concurrency:
+      group: zos-ssh-build
+    steps:
+      - name: Checkout
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          submodules: true
+
+      - name: Prepare
+        run: ./autogen.sh
+
+      - name: Build & test
+        env:
+          ZOS_HOST: ${{ secrets.ZOS_HOST }}
+          ZOS_PORT: ${{ secrets.ZOS_PORT }}
+          ZOS_PRIVATE_KEY: ${{ secrets.ZOS_PRIVATE_KEY }}
+          ZOS_KNOWN_HOSTS: ${{ secrets.ZOS_KNOWN_HOSTS }}
+        run: |
+          (umask 0077 && printenv ZOS_PRIVATE_KEY > id_rsa_zos)
+          mkdir -p ~/.ssh
+          printenv ZOS_KNOWN_HOSTS > ~/.ssh/known_hosts
+
+          tar czf ../pcre2-build.tar.gz --exclude=.git .
+          mv ../pcre2-build.tar.gz .
+
+          scp -i id_rsa_zos -P "$ZOS_PORT" pcre2-build.tar.gz "$ZOS_HOST:/data/"
+          ssh -i id_rsa_zos -p "$ZOS_PORT" "$ZOS_HOST" /data/zopen/usr/local/bin/bash -c \
+            'export _BPXK_AUTOCVT=ON;
+            export _CEE_RUNOPTS="FILETAG(AUTOCVT,AUTOTAG) POSIX(ON)";
+            export _TAG_REDIR_ERR=txt;
+            export _TAG_REDIR_IN=txt;
+            export _TAG_REDIR_OUT=txt;
+            . /data/zopen/etc/zopen-config;
+            set -e;
+            set -x;
+            cd /data;
+            rm -rf pcre2-build;
+            mkdir pcre2-build;
+            gtar xzf pcre2-build.tar.gz -C pcre2-build;
+            cd pcre2-build;
+            chtag -R -tc ISO8859-1 .;
+            MAKE=gmake CC=xlc ./configure --enable-ebcdic --disable-unicode;
+            gmake;
+            gmake check'
+
   distcheck:
     name: Build & verify distribution
     runs-on: ubuntu-latest

Makefile.am

@@ -235,6 +235,7 @@ lib_LTLIBRARIES =
 # TESTS is for binary unit tests, check_SCRIPTS for script-based tests
 
 TESTS =
+XFAIL_TESTS =
 check_SCRIPTS =
 dist_noinst_SCRIPTS =
 
@@ -725,6 +726,9 @@ if WITH_PCRE2_8
 TESTS += RunGrepTest
 EXTRA_DIST += RunGrepTest.bat
 dist_noinst_SCRIPTS += RunGrepTest
+if WITH_EBCDIC
+XFAIL_TESTS += RunGrepTest
+endif # WITH_EBCDIC
 endif # WITH_PCRE2_8
 
 ## Distribute all the test data files

RunTest

@@ -130,9 +130,11 @@ fi
 
 # Set up a suitable "diff" command for comparison. Some systems
 # have a diff that lacks a -u option. Try to deal with this.
+# Use gdiff if available.
 
 cf="diff"
-diff -u /dev/null /dev/null 2>/dev/null && cf="diff -u"
+gdiff /dev/null /dev/null 2>/dev/null && cf="gdiff"
+$cf -u /dev/null /dev/null 2>/dev/null && cf="$cf -u"
 
 # Find the test data
 

configure.ac

@@ -72,6 +72,9 @@ m4_ifdef([AM_PROG_AR], [AM_PROG_AR])
 # Check for a 64-bit integer type
 AC_TYPE_INT64_T
 
+# Check for xlc which has some special (broken/non-standard) behaviour on z/OS.
+PCRE2_ZOS_FIXES
+
 AC_PROG_INSTALL
 LT_INIT([win32-dll])
 AC_PROG_LN_S

m4/pcre2_zos.m4

@@ -0,0 +1,50 @@
+dnl Tests whether the compiler requires an additional flag in order to fail on
+dnl undefined headers.
+
+dnl The concept of setting this commandline flag was learned from patches and
+dnl mailing list discussions of the gnulib and gawk projects (credit to
+dnl Bruno Haible).
+
+AC_DEFUN([PCRE2_ZOS_FIXES],
+[
+  AC_CACHE_CHECK([for OS/390 (z/OS)], [pcre2_cv_os390],
+    [if test "`uname`" = "OS/390"; then
+       pcre2_cv_os390=yes
+     else
+       pcre2_cv_os390=no
+     fi])
+  if test "$pcre2_cv_os390" = "yes"; then
+    AC_CACHE_CHECK([whether the compiler supports -qhaltonmsg=CCN3296], [pcre2_cv_xlc_qhaltonmsg_support],
+      [save_CFLAGS="$CFLAGS"
+      CFLAGS="$CFLAGS -qhaltonmsg=CCN3296"
+      AC_COMPILE_IFELSE([AC_LANG_PROGRAM([])],
+                        [pcre2_cv_xlc_qhaltonmsg_support=yes],
+                        [pcre2_cv_xlc_qhaltonmsg_support=no])
+      CFLAGS="$save_CFLAGS"
+      ])
+
+    AC_CACHE_CHECK([whether non-existent headers fail the compile], [pcre2_cv_xlc_nonexistent_fatal],
+      [AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <thereshouldbenoheader.h>]])],
+                         [pcre2_cv_xlc_nonexistent_fatal=no],
+                         [pcre2_cv_xlc_nonexistent_fatal=yes])
+      ])
+
+    if test "$pcre2_cv_xlc_nonexistent_fatal" = "no" && test "$pcre2_cv_xlc_qhaltonmsg_support" = "yes"; then
+      AC_CACHE_CHECK([whether -qhaltonmsg=CCN3296 fixes the non-existent-header issue], [pcre2_cv_xlc_qhaltonmsg_fixes],
+        [save_CFLAGS="$CFLAGS"
+        CFLAGS="$CFLAGS -qhaltonmsg=CCN3296"
+        AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <thereshouldbenoheader.h>]])],
+                          [pcre2_cv_xlc_qhaltonmsg_fixes=no],
+                          [pcre2_cv_xlc_qhaltonmsg_fixes=yes])
+        CFLAGS="$save_CFLAGS"
+        ])
+
+      if test "$pcre2_cv_xlc_qhaltonmsg_fixes" = "no"; then
+        AC_MSG_ERROR([-qhaltonmsg=CCN3296 not effective on non-existent headers])
+      fi
+
+      CFLAGS="$CFLAGS -qhaltonmsg=CCN3296"
+    fi
+
+  fi
+])

maint/manifest-tarball

@@ -295,6 +295,7 @@ drwxr-xr-x tarball-dir/pcre2-SNAPSHOT/m4
 -rw-r--r-- tarball-dir/pcre2-SNAPSHOT/m4/ltversion.m4
 -rw-r--r-- tarball-dir/pcre2-SNAPSHOT/m4/lt~obsolete.m4
 -rw-r--r-- tarball-dir/pcre2-SNAPSHOT/m4/pcre2_visibility.m4
+-rw-r--r-- tarball-dir/pcre2-SNAPSHOT/m4/pcre2_zos.m4
 -rwxr-xr-x tarball-dir/pcre2-SNAPSHOT/missing
 -rw-r--r-- tarball-dir/pcre2-SNAPSHOT/pcre2-config.in
 -rwxr-xr-x tarball-dir/pcre2-SNAPSHOT/perltest.sh

src/pcre2_compile.c

@@ -55,11 +55,14 @@ POSSIBILITY OF SUCH DAMAGE.
 #else
 #define PRINTABLE(c) ((c) >= 32 && (c) < 127)
 #endif
-#define CHAR_OUTPUT(c)  (c)
-#define CHAR_INPUT(c)   (c)
+#define CHAR_OUTPUT(c)      (c)
+#define CHAR_OUTPUT_HEX(c)  (c)
+#define CHAR_INPUT(c)       (c)
+#define CHAR_INPUT_HEX(c)   (c)
 #include "pcre2_printint.c"
 #undef PRINTABLE
 #undef CHAR_OUTPUT
+#undef CHAR_OUTPUT_HEX
 #undef CHAR_INPUT
 #define DEBUG_CALL_PRINTINT
 #endif

src/pcre2_compile_class.c

@@ -1285,7 +1285,7 @@ while (TRUE)
       uint8_t posix_vertical[4] = { CHAR_LF, CHAR_VT, CHAR_FF, CHAR_CR };
       uint8_t posix_underscore = CHAR_UNDERSCORE;
       uint8_t *chars = NULL;
-      uint n = 0;
+      int n = 0;
 
       if (tabopt == 1) { chars = posix_vertical; n = 4; }
       else if (tabopt == 2) { chars = &posix_underscore; n = 1; }

src/pcre2_printint.c

@@ -120,10 +120,10 @@ if (utf)
 if (one_code_unit)
   {
   if (PRINTABLE(c))
-    fprintf(f, "%c", (char)CHAR_OUTPUT(c));
+    fprintf(f, "%c", CHAR_OUTPUT(c));
   else
     {
-    c = CHAR_OUTPUT(c);
+    c = CHAR_OUTPUT_HEX(c);
     if (c < 0x80) fprintf(f, "\\x%02x", c);
     else fprintf(f, "\\x{%02x}", c);
     }
@@ -219,7 +219,7 @@ while (*ptr != '\0')
   {
   uint32_t c = *ptr++;
   if (PRINTABLE(c)) fprintf(f, "%c", CHAR_OUTPUT(c));
-  else fprintf(f, "\\x{%x}", CHAR_OUTPUT(c));
+  else fprintf(f, "\\x{%x}", CHAR_OUTPUT_HEX(c));
   }
 }
 
@@ -230,7 +230,7 @@ for (; len > 0; len--)
   {
   uint32_t c = *ptr++;
   if (PRINTABLE(c)) fprintf(f, "%c", CHAR_OUTPUT(c));
-  else fprintf(f, "\\x{%x}", CHAR_OUTPUT(c));
+  else fprintf(f, "\\x{%x}", CHAR_OUTPUT_HEX(c));
   }
 }
 
@@ -481,22 +481,22 @@ if (negated)
 
 for (input = 0; input < 256; input++)
   {
-  i = CHAR_INPUT(input);
+  i = CHAR_INPUT_HEX(input);
   if ((map[i/8] & (1u << (i&7))) != 0)
     {
     int j, jinput;
     for (jinput = input; jinput+1 < 256; jinput++)
       {
-      j = CHAR_INPUT(jinput+1);
+      j = CHAR_INPUT_HEX(jinput+1);
       if ((map[j/8] & (1u << (j&7))) == 0) break;
       }
-    j = CHAR_INPUT(jinput);
+    j = CHAR_INPUT_HEX(jinput);
     if (i == CHAR_MINUS || i == CHAR_BACKSLASH ||
         i == CHAR_RIGHT_SQUARE_BRACKET ||
         (first && i == CHAR_CIRCUMFLEX_ACCENT))
       fprintf(f, "\\");
     if (PRINTABLE(i)) fprintf(f, "%c", CHAR_OUTPUT(i));
-    else fprintf(f, "\\x%02x", CHAR_OUTPUT(i));
+    else fprintf(f, "\\x%02x", CHAR_OUTPUT_HEX(i));
     first = FALSE;
     if (jinput > input)
       {
@@ -505,7 +505,7 @@ for (input = 0; input < 256; input++)
           j == CHAR_RIGHT_SQUARE_BRACKET)
         fprintf(f, "\\");
       if (PRINTABLE(j)) fprintf(f, "%c", CHAR_OUTPUT(j));
-      else fprintf(f, "\\x%02x", CHAR_OUTPUT(j));
+      else fprintf(f, "\\x%02x", CHAR_OUTPUT_HEX(j));
       }
     input = jinput;
     }