"sed -E" should work, no? '\t' is supported in extended regular expressions.
-DrD- > Hi Erik, > I applied the patch in a previously failing build environment and it > appears to do the business. > The rmic -iiop is producing its stubs and tie classes. > > thanks for the fix. > > regards > Mark > On 28/04/2014 12:57, Erik Joelsson wrote: >> Please review this small patch which is correcting the "clean" properties >> feature in the SetupJavaCompilation macro. On Macosx, sed does not >> understand '\t' which we use to match tab characters. The consequence of >> this is that the character 't' gets removed at beginning and end of lines >> when cleaned on Macosx. The solution is to explicitly add a tab character >> instead. >> >> Tested that it fixes the problem on Macosx and that nothing changes on other >> platforms. >> >> Bug: https://bugs.openjdk.java.net/browse/JDK-8041265 >> Patch inline: >> diff -r ab55a18a95e1 make/common/JavaCompilation.gmk >> --- a/make/common/JavaCompilation.gmk >> +++ b/make/common/JavaCompilation.gmk >> @@ -351,7 +351,8 @@ >> # 3. Delete all lines starting with #. >> # 4. Delete empty lines. >> # 5. Append lines ending with \ with the next line. >> -# 6. Remove leading and trailing white space. >> +# 6. Remove leading and trailing white space. Note that tabs must be >> explicit >> +# as sed on macosx does not understand '\t'. >> # 7. Replace the first \= with just =. >> # 8. Finally it's all sorted to create a stable output. >> # >> @@ -370,7 +371,7 @@ >> | $(SED) -f "$(SRC_ROOT)/make/common/support/unicode2x.sed" \ >> | $(SED) -e '/^#/d' -e '/^$$$$/d' \ >> -e :a -e '/\\$$$$/N; s/\\\n//; ta' \ >> - -e 's/^[ \t]*//;s/[ \t]*$$$$//' \ >> + -e 's/^[ ]*//;s/[ ]*$$$$//' \ >> -e 's/\\=/=/' | LC_ALL=C $(SORT) > $$@ >> $(CHMOD) -f ug+w $$@ >> >
