Re: [Monotone-devel] merge bug? - silent data change

[Julio M. Merino Vidal]

On Aug 26, 2007, at 7:19 PM, Markus Schiltknecht wrote:

> Hello Pavel,
>
> Pavel Cahyna wrote:
> > it can be obtained by mtn pull ufs-trans.mtn-host.prjek.net.
> > look at the diff of sys/dev/wscons/wsdisplay_vcons.c
> > (http://mtn-host.prjek.net/viewmtn/ufs-trans/revision/diff/ 
> > 6eb5489725a8c1b10bab48b8341fc5b3e57287a4/with/ 
> > 951b0672754af7e3272775d548c04e3659089bd6/sys/dev/wscons/ 
> > wsdisplay_vcons.c)
> > it is obviously nonsensical. I think that there is no reason to  
> > produce
> > such a difference so it is apparently a merge bug.
>
> The diff looks somewhat like the problem I had until recently. It's  
> been covered by test_a_merge_8 and I've (hopefully) fixed it just  
> recently.
>
> Do you have a change of trying the merge with a bleeding-edge (i.e.  
> newest checkout) monotone version?
>
> Unfortunately, I'm currently on vacation, otherwise I'd give it a try.
>
> Please let us know if the newish monotone produces a more sane merge.

I have tried the merge again with a recent monotone on Pavel's behalf  
and it seems to produce a sane revision.

Comparing the bad merge with the new one delivers this diff:

calypso:/tmp> ~/local/bin/mtn --db=test.mtn diff -r  
951b0672754af7e3272775d548c04e3659089bd6 -r 19e17
mtn: expanded selector '19e17' -> 'i:19e17'
mtn: expanding selection '19e17'
mtn: expanded to '19e1714b623e32ac5fa010300412cb4d3df2bf8c'
#
#
# patch "sys/dev/wscons/wsdisplay_vcons.c"
#  from [d343f9bf977d536b7a3e856213a29f3c9d48dbe6]
#    to [350ae716c907f392c9b6e450cc52649da7c885e3]
#
============================================================
--- sys/dev/wscons/wsdisplay_vcons.c     
d343f9bf977d536b7a3e856213a29f3c9d48dbe6
+++ sys/dev/wscons/wsdisplay_vcons.c     
350ae716c907f392c9b6e450cc52649da7c885e3
@@ -576,11 +576,11 @@ vcons_erasecols_buffer(void *cookie, int
        int offset;
        offset = scr->scr_offset_to_zero;
-#ifdef WSDISPLAY_SCROLLSUPPORT
-       int offset;
-       offset = scr->scr_offset_to_zero;
-
        for (i = start; i < end; i++) {
+               scr->scr_attrs[offset + i] = fillattr;
+               scr->scr_chars[offset + i] = 0x20;
+       }
+#else
        for (i = start; i < end; i++) {
                scr->scr_attrs[i] = fillattr;
                scr->scr_chars[i] = 0x20;

Which feels correct.

--
Julio M. Merino Vidal <address@hidden>