Skip to content

Latest commit

 

History

History
112 lines (109 loc) · 5.58 KB

todo.org

File metadata and controls

112 lines (109 loc) · 5.58 KB

TASKS

BUG Replacing dirs with files and vice versa crashes

see http://p42svn.tigris.org/issues/show_bug.cgi?id=8 test case checked in: r176 | trentf | 2013-05-21 12:28:59 -0700 (Tue, 21 May 2013) | 1 line r175 | trentf | 2013-05-21 12:28:36 -0700 (Tue, 21 May 2013) | 1 line

Timezone handling of checkin dates

  • State “DONE” from “” [2012-03-27 Tue 11:30]

It seems p42svn takes the dates from P4 as utc sample depot Change 12000 by perfmerge++@perfmerge++ on 2011/05/16 15:03:40 in dump file: Revision-number: 462 2011-05-16T15:03:40.000000Z p4_get_change_details() is where this needs to be adjusted add “utc” to time2str, otherwise it uses local time testcase in timezone.t

non utf8 chars in log messages

An attempt at this was done in svn_props2string() but I couldn’t get the byte counts right Rajesh’s patch fixes this

case insensitivity in depot names: //doc vs //Doc

  • State “DONE” from “BUG” [2012-04-11 Wed 10:28]

only an issue on case insensitive servers both labels and branches need to do this Added code in p4_get_label_files() and is_wanted_file() which does case insensitive matches if –fix-case is used

case mapping support with –changes

  • State “DONE” from “BUG” [2012-04-11 Wed 10:28]

refactor casemapping code and use it to process the –existing-files list

ETA calculations wrong on incremental imports

  • State “DONE” from “NEXT” [2012-12-21 Fri 11:02]

Improve –verify

check files changed in each rev Update: this has been implemented, but I’m not sure it’s 100% right only checks for changes, not file contents… that would be as slow as a migration! should add properties to repository for verified revs

Add better statistics

count revs, actions, content (+cache)

Fix handling of unintelligible characters

% escapes was a bad idea Perhaps use the new option to encode which takes a code ref for dealing with bogus chars… but that requires Perl 5.8.8 or newer use an = escape, a la quoted-printable encoding

Better die() handling

http://p42svn.tigris.org/ds/viewMessage.do?dsMessageId=2918677&dsForumId=4895

provide a way to stop p42svn and p42svnsync cleanly

  • State “DONE” from “NEXT” [2013-01-02 Wed 12:35]

p42svn should accept a signal telling it to exit after finishing this rev. Should be done for INT, HUP, ?? p42svnsync should pass that signal to p42svn and wait for them to finish. should multiple interupts kill it anyway, or just make them do a kill -9? decided to use a marker file with the PID on it (by default) p42svnsync now tries to deal with INT and HUP by killing svnadmin first.

BUG panic: sv_setpvn

User report: I had trouble getting past “panic: sv_setpvn called with negative strlen” when it was running the line: “$result = $p4->Run(‘print’,$filename);”, so I simply replaced the line with direct invocation of “p4 print –q $filename” and all seemed fine.

Do you have reason to believe that doing so is not identical to doing “Run(‘print’, $filename)” ?

FYI the below is the related thread that someone else started where I provided my own input: http://forums.perforce.com/index.php?/topic/756-p4perlc-api-issues-with-run/page__gopid__4142#entry4142

SOMEDAY optimize branch, integrate and label to copy entire tree

for example a branch action will list every file in the tree but in svn that can be a copy of the top of the tree Not sure if this is even possible with the data at hand: the branch change list is based on the versions (#) not change numbers (@) So I don’t know if there’s any way to determine which change a given branch or label is without doing even more queries against P4, which is where the performance suffers.

fail on outside scope copies

if the origin of a label, branch or integrate is not within scope, it is blindly turned into a copy. however, it is possible that these are signs of bad set of branch specs so, should we issue warnings, or maybe error out given a cmd line option?

Test suite

Set up a self contained depot instead of the stock sample file

needs to be a case insensitive server check in the checkpointed tar ball

finish converting test cases to perl-based tests

  • State “DONE” from “NEXT” [2013-01-02 Wed 15:57]

Verify that conversion test is correct

that is, changing file types, like link replaced with a dir, &c. Could someone who knows P4 review?

test cases for case insensitivity

how to set up test cases, another depot with -C1? or switch existing server?

test cases for weird characters

is rawcharset.t sufficient?

test case for integrating from other depots, including removed ones

test case for purge

more test cases for move operations

currently the assumption is that move/add and move/delete operations will occur together in a given changeset, and than simply doing the former before the latter will ensure the operations get converted properly. Is this assumption wrong

test cases for labels

labels can show delete, move/delete and purge operations. these get excluded at conversion time. the test depot has no such labels… how do they get generated?