aboutsummaryrefslogtreecommitdiff
path: root/gold
diff options
context:
space:
mode:
Diffstat (limited to 'gold')
-rw-r--r--gold/ChangeLog5
-rw-r--r--gold/object.cc9
2 files changed, 10 insertions, 4 deletions
diff --git a/gold/ChangeLog b/gold/ChangeLog
index fdedd86..819f250 100644
--- a/gold/ChangeLog
+++ b/gold/ChangeLog
@@ -1,3 +1,8 @@
+2011-07-05 Ian Lance Taylor <iant@google.com>
+
+ * object.cc (Sized_relobj_file::include_section_group): Add
+ information to comment about signature location.
+
2011-07-02 Ian Lance Taylor <iant@google.com>
PR gold/12957
diff --git a/gold/object.cc b/gold/object.cc
index d3835df..6f68ed3 100644
--- a/gold/object.cc
+++ b/gold/object.cc
@@ -748,10 +748,11 @@ Sized_relobj_file<size, big_endian>::include_section_group(
// just like ordinary sections.
elfcpp::Elf_Word flags = elfcpp::Swap<32, big_endian>::readval(pword);
- // Look up the group signature, which is the name of a symbol. This
- // is a lot of effort to go to to read a string. Why didn't they
- // just have the group signature point into the string table, rather
- // than indirect through a symbol?
+ // Look up the group signature, which is the name of a symbol. ELF
+ // uses a symbol name because some group signatures are long, and
+ // the name is generally already in the symbol table, so it makes
+ // sense to put the long string just once in .strtab rather than in
+ // both .strtab and .shstrtab.
// Get the appropriate symbol table header (this will normally be
// the single SHT_SYMTAB section, but in principle it need not be).