#2805: Ticket view sometimes silently drops information
--------------------------------+-------------------------------------------
Reporter: [EMAIL PROTECTED] | Owner: jonas
Type: defect | Status: new
Priority: high | Milestone: 0.10
Component: ticket system | Version: devel
Severity: major | Resolution:
Keywords: |
--------------------------------+-------------------------------------------
Old description:
> I've got a Ticket with a changelog (as returned by get_changelog) that
> looks like this:
>
> [(1139910447,
> 'miro',
> 'comment',
> '',
> '<long important comment snipped>'),
> (1139910447, 'miro', 'comment', '', ''),
> (1139910447, 'miro', 'attachment', '', 'sshserver.py')]
>
> The 1st 'comment' entry comes from an actual comment on the ticket. The
> 2nd 'comment' entry comes from the description of the attachment on the
> attachment table - ie, it is a row returned by the final SELECT clause in
> the statement which retrieves the changelog (line ~288 in
> trac/ticket/model.py). Since the timestamp and author on these two rows
> are the same, when the web ui builds up the change data set for the
> rendering pass (line ~420 in trac/ticket/web_ui.py), it does not create a
> new `changes' entry, but merely overwrites the first comment with the
> second.
>
> It seems as though the schema is missing a reliable way to associate
> multiple pieces of change information with each other. The ad hoc
> time/author mechanism cannot be counted on to work in all cases.
>
> I've patched my local version of trac like this in order to temporarily
> work around the problem:
>
> {{{
> === trac/ticket/web_ui.py
> ==================================================================
> --- trac/ticket/web_ui.py (revision 2048)
> +++ trac/ticket/web_ui.py (local)
> @@ -418,7 +418,7 @@
> curr_date = 0
> changes = []
> for date, author, field, old, new in changelog:
> - if date != curr_date or author != curr_author:
> + if date != curr_date or author != curr_author or (field in
> changes[-1]):
> changes.append({
> 'date': util.format_datetime(date),
> 'author': author,
> }}}
New description:
I've got a Ticket with a changelog (as returned by get_changelog) that
looks like this:
{{{
[(1139910447,
'miro',
'comment',
'',
'<long important comment snipped>'),
(1139910447, 'miro', 'comment', '', ''),
(1139910447, 'miro', 'attachment', '', 'sshserver.py')]
}}}
The 1st 'comment' entry comes from an actual comment on the ticket. The
2nd 'comment' entry comes from the description of the attachment on the
attachment table - ie, it is a row returned by the final SELECT clause in
the statement which retrieves the changelog (line ~288 in
trac/ticket/model.py). Since the timestamp and author on these two rows
are the same, when the web ui builds up the change data set for the
rendering pass (line ~420 in trac/ticket/web_ui.py), it does not create a
new `changes' entry, but merely overwrites the first comment with the
second.
It seems as though the schema is missing a reliable way to associate
multiple pieces of change information with each other. The ad hoc
time/author mechanism cannot be counted on to work in all cases.
I've patched my local version of trac like this in order to temporarily
work around the problem:
{{{
=== trac/ticket/web_ui.py
==================================================================
--- trac/ticket/web_ui.py (revision 2048)
+++ trac/ticket/web_ui.py (local)
@@ -418,7 +418,7 @@
curr_date = 0
changes = []
for date, author, field, old, new in changelog:
- if date != curr_date or author != curr_author:
+ if date != curr_date or author != curr_author or (field in
changes[-1]):
changes.append({
'date': util.format_datetime(date),
'author': author,
}}}
--
Ticket URL: <http://projects.edgewall.com/trac/ticket/2805>
The Trac Project <http://trac.edgewall.com/>
_______________________________________________
Trac-Tickets mailing list
[email protected]
http://lists.edgewall.com/mailman/listinfo/trac-tickets
