* scripts/abilist.awk: Produce a more compact format, divided into

	stanzas for each version set, the set name listed only once.
	* scripts/extract-abilist.awk: New file.
	* scripts/merge-abilist.awk: New file.
	* Makerules (check-abi-%, update-abi-%): New pattern rules.
	(update-abi, check-abi): New targets.
	* Makefile (+subdir_targets): Add subdir_{check,update}-abi.

	* Makerules (%.symlist): Use LC_ALL=C when running awk script.

3 files changed, 233 insertions, 10 deletions

diff --git a/scripts/abilist.awk b/scripts/abilist.awk
index 9781834..346527e 100644
--- a/scripts/abilist.awk
+++ b/scripts/abilist.awk
@@ -1,8 +1,6 @@
 # This awk script processes the output of objdump --dynamic-syms
 # into a simple format that should not change when the ABI is not changing.
 
-BEGIN { outpipe = "sort" }
-
 # Normalize columns.
 /^[0-9a-fA-F]+      / { sub(/      /, "  -   ") }
 
@@ -13,11 +11,11 @@ $4 == "*UND*" { next }
 $2 == "l" { next }
 
 $2 == "g" || $2 == "w" && NF == 7 {
-  weak = ($2 == "w") ? "weak" : "strong";
+  weak = $2;
   type = $3;
   size = $5;
   sub(/^0*/, "", size);
-  size = "0x" size;
+  size = " 0x" size;
   version = $6;
   symbol = $7;
   gsub(/[()]/, "", version);
@@ -25,24 +23,36 @@ $2 == "g" || $2 == "w" && NF == 7 {
   if (version == "GLIBC_PRIVATE") next;
 
   if (type == "D" && $4 == ".tbss") {
-    print symbol, version, weak, "TLS", size | outpipe;
+    type = "T";
   }
   else if (type == "D" && $4 == ".opd") {
-    print symbol, version, weak, "FDESC" | outpipe;
+    type = "O";
+    size = "";
   }
   else if (type == "DO" && $4 == "*ABS*") {
-    print symbol, version, weak, "ABS" | outpipe;
+    type = "A";
+    size = "";
   }
   else if (type == "DO") {
-    print symbol, version, weak, "DATA", size | outpipe;
+    type = "D";
   }
   else if (type == "DF") {
-    print symbol, version, weak, "FUNC" | outpipe;
+    type = "F";
+    size = "";
   }
   else {
-    print symbol, version, weak, "UNEXPECTED", type, $4, $5;
+    print symbol, version, weak, "?", type, $4, $5;
+    next;
   }
 
+  desc = " " symbol " " (weak == "w" ? tolower(type) : type) size;
+
+  if (version in versions) {
+    versions[version] = versions[version] "\n" desc;
+  }
+  else {
+    versions[version] = desc;
+  }
   next;
 }
 
@@ -52,3 +62,43 @@ NF == 0 || /DYNAMIC SYMBOL TABLE/ || /file format/ { next }
 {
   print "Don't grok this line:", $0
 }
+
+END {
+  nverlist = 0;
+  for (version in versions) {
+    if (nverslist == 0) {
+      verslist = version;
+      nverslist = 1;
+      continue;
+    }
+    split(verslist, s, "\n");
+    if (version < s[1]) {
+      verslist = version;
+      for (i = 1; i <= nverslist; ++i) {
+	verslist = verslist "\n" s[i];
+      }
+    }
+    else {
+      verslist = s[1];
+      for (i = 2; i <= nverslist; ++i) {
+	if (version < s[i]) break;
+	verslist = verslist "\n" s[i];
+      }
+      verslist = verslist "\n" version;
+      for (; i <= nverslist; ++i) {
+	verslist = verslist "\n" s[i];
+      }
+    }
+    ++nverslist;
+  }
+
+  split(verslist, order, "\n");
+  for (i = 1; i <= nverslist; ++i) {
+    version = order[i];
+
+    print version;
+    outpipe = "sort";
+    print versions[version] | outpipe;
+    close(outpipe);
+  }
+}
diff --git a/scripts/extract-abilist.awk b/scripts/extract-abilist.awk
new file mode 100644
index 0000000..c5b76e1
--- /dev/null
+++ b/scripts/extract-abilist.awk
@@ -0,0 +1,44 @@
+# awk script to extract a config-specific .symlist file from a merged file.
+# This must be passed run with awk -v config=TUPLE to specify the configuration
+# tuple we will match.  The merged file contains stanzas in the form:
+#	GLIBC_x.y regexp...
+#	 function F
+#	 variable D 0x4
+# Each regexp is matched against TUPLE, and only matching stanzas go
+# into the output, with the regexp list removed.  The result matches the
+# original .symlist file from abilist.awk that was fed into merge-abilist.awk.
+
+BEGIN {
+  outpipe = "";
+}
+
+/^ / { if (!ignore) print | outpipe; next; }
+
+{
+  for (i = 2; i <= NF; ++i) {
+    regex = "^" $i "$";
+    if (match(config, regex) != 0) {
+      if ($1 != version) {
+	if (outpipe != "") {
+	  close(outpipe);
+	}
+	version = $1;
+	print version;
+	outpipe = "sort";
+      }
+      ignore = 0;
+      next;
+    }
+  }
+  ignore = 1;
+  next;
+}
+
+END {
+  if (outpipe == "") {
+    print "No stanza matched", config > "/dev/stderr";
+    exit 2;
+  }
+  else
+    close(outpipe);
+}
diff --git a/scripts/merge-abilist.awk b/scripts/merge-abilist.awk
new file mode 100644
index 0000000..016debc
--- /dev/null
+++ b/scripts/merge-abilist.awk
@@ -0,0 +1,129 @@
+# awk script to merge a config-specific .symlist file with others.
+# The input files should be an existing .abilist file, and a .symlist file.
+# This must be passed run with awk -v config=REGEXP to specify a regexp
+# matching configuration tuples for which the .symlist input defines an ABI.
+# The result merges all duplicate occurrences of any symbol in a version set
+# into a stanza listing the regexps matching configurations that contain it.
+
+/^[^ ]/ {
+  if (NF < 2 && config == "") {
+    print "BAD LINE:", $0 > "/dev/stderr";
+    exit 2;
+  }
+
+  if (NF < 2) {
+    current = $1 ":" config;
+  }
+  else {
+    current = $1 ":" $2;
+    for (i = 3; i <= NF; ++i) {
+      current = current "," $1 ":" $i;
+    }
+  }
+
+  next;
+}
+
+{
+  if ($0 in seen) {
+    seen[$0] = seen[$0] "\n" current;
+  }
+  else {
+    seen[$0] = current;
+  }
+
+  next;
+}
+
+END {
+  for (line in seen) {
+    split(seen[line], setlist, "\n");
+    for (i in setlist) {
+      split(setlist[i], configs, ",");
+      for (j in configs) {
+	split(configs[j], temp, ":");
+	version = temp[1];
+	conf = temp[2];
+
+	if ((version,conf) in have) continue;
+	have[version,conf] = 1;
+
+	if (version in confs) {
+	  split(confs[version], c, " ");
+	  if (conf < c[1]) {
+	    confs[version] = conf;
+	    for (k = 1; k <= nconfs[version]; ++k) {
+	      confs[version] = confs[version] " " c[k];
+	    }
+	  }
+	  else {
+	    confs[version] = c[1];
+	    for (k = 2; k <= nconfs[version]; ++k) {
+	      if (conf < c[k]) break;
+	      confs[version] = confs[version] " " c[k];
+	    }
+	    confs[version] = confs[version] " " conf;
+	    for (; k <= nconfs[version]; ++k) {
+	      confs[version] = confs[version] " " c[k];
+	    }
+	  }
+	  ++nconfs[version];
+	}
+	else {
+	  confs[version] = conf;
+	  nconfs[version] = 1;
+	}
+      }
+    }
+    for (idx in have) delete have[idx];
+
+    for (version in confs) {
+      idx = version " " confs[version];
+      if (idx in final) {
+	final[idx] = final[idx] "\n" line;
+      }
+      else {
+	final[idx] = line;
+      }
+      delete confs[version];
+      delete nconfs[version];
+    }
+  }
+
+  nstanzas = 0;
+  for (stanza in final) {
+    if (nstanzas == 0) {
+      stanzas = stanza;
+      nstanzas = 1;
+      continue;
+    }
+    split(stanzas, s, "\n");
+    if (stanza < s[1]) {
+      stanzas = stanza;
+      for (i = 1; i <= nstanzas; ++i) {
+	stanzas = stanzas "\n" s[i];
+      }
+    }
+    else {
+      stanzas = s[1];
+      for (i = 2; i <= nstanzas; ++i) {
+	if (stanza < s[i]) break;
+	stanzas = stanzas "\n" s[i];
+      }
+      stanzas = stanzas "\n" stanza;
+      for (; i <= nstanzas; ++i) {
+	stanzas = stanzas "\n" s[i];
+      }
+    }
+    ++nstanzas;
+  }
+
+  split(stanzas, order, "\n");
+  for (i = 1; i <= nstanzas; ++i) {
+    stanza = order[i];
+    print stanza;
+    outpipe = "sort";
+    print final[stanza] | outpipe;
+    close(outpipe);
+  }
+}