Community
Participate
Working Groups
Using I20090811-0800 but it also happens in 3.5.0 I discovered that using HEAD the formatter was unstable while performing the bug 286668 comment 2 test case: public class Test { void foo() { StringBuilder builder = new StringBuilder(); builder.append("abc").append("def").append("ghi").append("jkl").append("mno") .append("pqr").append("stu").append("vwx").append("yz"); } } is formatted as follow: public class Test { void foo() { StringBuilder builder = new StringBuilder(); builder.append("abc").append( "def").append("ghi") .append("jkl").append( "mno") .append("pqr").append( "stu").append( "vwx").append( "yz"); } } That looks not too bad, but if I format again this code, expecting that nothing changes, I get the following wrong output: public class Test { void foo() { StringBuilder builder = new StringBuilder(); builder.append("abc").append( "def").append("ghi") .append("jkl").append( "mno") .append("pqr").append( "stu").append( "vwx").append( "yz"); } }
I'm currently trying to figure out how fixing this problematic issue...
Created attachment 144846 [details] Proposed patch A tricky one... This patch should hopefully fix most of the issue around the 'Never join lines' preferences. It removes fix done for bug 286601 which was not enough solid to cover all cases. The current strategy is now to analyze the current indentation which exist on the broken line and apply the same after having indented the current alignment. This allow to get rid of the path currently followed in the CodeFormatterVisitor when trying to preserve the broken line, hence make this fix more general than the previous one. Of course it also fixes, and that was obviously its main goal, the stability problem of the formatter...
Olivier, could you please review and let me know if you agree with the strategy applied in this patch? Thanks
Created attachment 145241 [details] New proposed patch This patch fixes regression observed while running massive formatter tests. Note that with this new patch all the changes done in patch for bug 286601 have been reverted (especially the change done in bug 286601 comment 6). With this new patch we're back to the 3.6M1 and 3.5.0 behavior, which was unfortunately not the case since bug 286601 fix was released...
Released for 3.6M2 in HEAD stream.
(In reply to comment #4) > > With this new patch we're back to the 3.6M1 and 3.5.0 behavior, which was > unfortunately not the case since bug 286601 fix was released... > May be this was not crystal clear... I meant that with this new patch, the bug 286601 is still fixed but without the regressions introduced by the patch initially released to fix it.
Verified for 3.6M2 using I20090914-1800
Verified.
*** Bug 286601 has been marked as a duplicate of this bug. ***
