On 2016-10-11 11:07, Erik Joelsson wrote:
Looks ok.
What happens in the build if a malformed manifest file is introduced
again? Will it fail?
If I remember correctly, the result of that would not be a failure
during the build process, but instead a malformed manifest file will be
generated in the output. I presume this would be detected by testing.
But a new malformed file is unlikely, that this one survived as long was
"thanks" to the old hotspot build that was specially written to handle it.
/Magnus
/Erik
On 2016-10-11 10:52, Magnus Ihse Bursie wrote:
Due to a piece of broken internal code, a workaround was added to
JarArchive to handle misformed manifest files. The root cause have
now been fixed, and the workaround should be removed and the original
code restored.
Bug: https://bugs.openjdk.java.net/browse/JDK-8158181
Patch inline:
diff --git a/make/common/JarArchive.gmk b/make/common/JarArchive.gmk
--- a/make/common/JarArchive.gmk
+++ b/make/common/JarArchive.gmk
@@ -242,7 +242,7 @@
# potential changes.
$$(if $$(filter $$($1_VARDEPS_FILE) $$($1_MANIFEST), $$?), \
$$(if $$($1_MANIFEST), \
- $(SED) -e '$(DOLLAR)$(DOLLAR)a\' $$($1_MANIFEST) >
$$($1_MANIFEST_FILE) $$(NEWLINE) \
+ $(CP) $$($1_MANIFEST) $$($1_MANIFEST_FILE) $$(NEWLINE) \
, \
$(RM) $$($1_MANIFEST_FILE) && $(TOUCH)
$$($1_MANIFEST_FILE) $$(NEWLINE)) \
$$(if $$($1_JARMAIN), \
/Magnus