From f378a5a316682247bc9385594706518461eaacde Mon Sep 17 00:00:00 2001
From: Phlosioneer <mattmdrr2@gmail.com>
Date: Sun, 3 Mar 2019 15:02:26 -0500
Subject: [PATCH] Fix symbol overcounting, add partial symbols

---
 .travis/calcrom/calcrom.pl | 57 +++++++++++++++++++++++++++++++++-----
 1 file changed, 50 insertions(+), 7 deletions(-)

diff --git a/.travis/calcrom/calcrom.pl b/.travis/calcrom/calcrom.pl
index 354f9bcee5..9c48679290 100755
--- a/.travis/calcrom/calcrom.pl
+++ b/.travis/calcrom/calcrom.pl
@@ -31,28 +31,66 @@ while (my $line = <$file>)
     }
 }
 
-# It sucks that we have to objdump twice, but I can't figure out how to get
-# stdin working for subcommands in perl while still having a timeout.
+# Note that the grep filters out all branch labels. It also requires a minimum
+# line length of 5, to filter out a ton of generated symbols (like AcCn). No
+# settings to nm seem to remove these symbols. Finally, nm prints out a separate
+# entry for whenever a name appears in a file, not just where it's defined. uniq
+# removes all the duplicate entries.
+#
+#
+# You'd expect this to take a while, because of uniq. It runs in under a second,
+# though. Uniq is pretty fast!
+my $base_cmd = "nm pokeemerald.elf | awk '{print \$3}' | grep '^[^_].\\{4\\}' | uniq";
+
+# This looks for Unknown_, Unknown_, or sub_, followed by just numbers. Note that
+# it matches even if stuff precedes the unknown, like sUnknown/gUnknown.
+my $undoc_cmd = "grep '[Uu]nknown_[0-9a-fA-F]*\\|sub_[0-9a-fA-F]*'";
+
+# This looks for every symbol with an address at the end of it. Some things are
+# given a name based on their type / location, but still have an unknown purpose.
+# For example, FooMap_EventScript_FFFFFFF.
+my $partial_doc_cmd = "grep '[0-9a-fA-F]\\{6,7\\}'";
+
+my $count_cmd = "wc -l";
+
+# It sucks that we have to run this three times, but I can't figure out how to get
+# stdin working for subcommands in perl while still having a timeout. It's decently
+# fast anyway.
 my $total_syms_as_string;
 (run (
-    command => "arm-none-eabi-nm pokeemerald.elf | wc -l",
+    command => "$base_cmd | $count_cmd",
     buffer => \$total_syms_as_string,
-    timeout => 30
+    timeout => 60
 ))
     or die "ERROR: Error while getting all symbols: $?";
 
 my $undocumented_as_string;
 (run (
-    command => "arm-none-eabi-nm pokeemerald.elf | grep \"Unknown_\\|sub_\" | wc -l",
+    command => "$base_cmd | $undoc_cmd | $count_cmd",
     buffer => \$undocumented_as_string,
-    timeout => 30
+    timeout => 60
 ))
     or die "ERROR: Error while filtering for undocumented symbols: $?";
 
+my $partial_documented_as_string;
+(run (
+    command => "$base_cmd | $partial_doc_cmd | $count_cmd",
+    buffer => \$partial_documented_as_string,
+    timeout => 60
+))
+    or die "ERROR: Error while filtering for partial symbols: $?";
+
+# Performing addition on a string converts it to a number. Any string that fails
+# to convert to a number becomes 0. So if our converted number is 0, but our string
+# is nonzero, then the conversion was an error.
 my $undocumented = $undocumented_as_string + 0;
 (($undocumented != 0) and ($undocumented_as_string ne "0"))
     or die "ERROR: Cannot convert string to num: '$undocumented_as_string'";
 
+my $partial_documented = $partial_documented_as_string + 0;
+(($partial_documented != 0) and ($partial_documented_as_string ne "0"))
+	or die "ERROR: Cannot convert string to num: '$partial_documented_as_string'";
+
 my $total_syms = $total_syms_as_string + 0;
 (($total_syms != 0) and ($total_syms_as_string ne "0"))
     or die "ERROR: Cannot convert string to num: '$total_syms_as_string'";
@@ -64,8 +102,12 @@ my $total = $src + $asm;
 my $srcPct = sprintf("%.4f", 100 * $src / $total);
 my $asmPct = sprintf("%.4f", 100 * $asm / $total);
 
-my $documented = $total_syms - $undocumented;
+# partial_documented is double-counting the unknown_* and sub_* symbols.
+$partial_documented = $partial_documented - $undocumented;
+
+my $documented = $total_syms - ($undocumented + $partial_documented);
 my $docPct = sprintf("%.4f", 100 * $documented / $total_syms);
+my $partialPct = sprintf("%.4f", 100 * $partial_documented / $total_syms);
 my $undocPct = sprintf("%.4f", 100 * $undocumented / $total_syms);
 
 print "$total total bytes of code\n";
@@ -74,4 +116,5 @@ print "$asm bytes of code in asm ($asmPct%)\n";
 print "\n";
 print "$total_syms total symbols\n";
 print "$documented symbols documented ($docPct%)\n";
+print "$partial_documented symbols partially documented ($partialPct%)\n";
 print "$undocumented symbols undocumented ($undocPct%)\n";