wiki:TracQuery

Trac Ticket Queries

In addition to reports, Trac provides support for custom ticket queries, used to display lists of tickets meeting a specified set of criteria.

To configure and execute a custom query, switch to the View Tickets module from the navigation bar, and select the Custom Query link.

Filters

When you first go to the query page the default filter will display tickets relevant to you:

  • If logged in then all open tickets it will display open tickets assigned to you.
  • If not logged in but you have specified a name or email address in the preferences then it will display all open tickets where your email (or name if email not defined) is in the CC list.
  • If not logged and no name/email defined in the preferences then all open issues are displayed.

Current filters can be removed by clicking the button to the left with the minus sign on the label. New filters are added from the pulldown lists at the bottom corners of the filters box ('And' conditions on the left, 'Or' conditions on the right). Filters with either a text box or a pulldown menu of options can be added multiple times to perform an or of the criteria.

You can use the fields just below the filters box to group the results based on a field, or display the full description for each ticket.

Once you've edited your filters click the Update button to refresh your results.

Clicking on one of the query results will take you to that ticket. You can navigate through the results by clicking the Next Ticket or Previous Ticket links just below the main menu bar, or click the Back to Query link to return to the query page.

You can safely edit any of the tickets and continue to navigate through the results using the Next/Previous/Back to Query links after saving your results. When you return to the query any tickets which were edited will be displayed with italicized text. If one of the tickets was edited such that it no longer matches the query criteria the text will also be greyed. Lastly, if a new ticket matching the query criteria has been created, it will be shown in bold.

The query results can be refreshed and cleared of these status indicators by clicking the Update button again.

Saving Queries

Trac allows you to save the query as a named query accessible from the reports module. To save a query ensure that you have Updated the view and then click the Save query button displayed beneath the results. You can also save references to queries in Wiki content, as described below.

Note: one way to easily build queries like the ones below, you can build and test the queries in the Custom report module and when ready - click Save query. This will build the query string for you. All you need to do is remove the extra line breaks.

Note: you must have the REPORT_CREATE permission in order to save queries to the list of default reports. The Save query button will only appear if you are logged in as a user that has been granted this permission. If your account does not have permission to create reports, you can still use the methods below to save a query.

You may want to save some queries so that you can come back to them later. You can do this by making a link to the query from any Wiki page.

[query:status=new|assigned|reopened&version=1.0 Active tickets against 1.0]

Which is displayed as:

Active tickets against 1.0

This uses a very simple query language to specify the criteria (see Query Language).

Alternatively, you can copy the query string of a query and paste that into the Wiki link, including the leading ? character:

[query:?status=new&status=assigned&status=reopened&group=owner Assigned tickets by owner]

Which is displayed as:

Assigned tickets by owner

Using the [[TicketQuery]] Macro

The TicketQuery macro lets you display lists of tickets matching certain criteria anywhere you can use WikiFormatting.

Example:

[[TicketQuery(version=0.6|0.7&resolution=duplicate)]]

This is displayed as:

#35
several nightly tests failing with context_id related errors
#44
Re: [MPICH2 Req #3768] Problem with MPICH2
#112
Last night's tests
#125
A problem about using mpd
#172
multiple definition?
#173
Resized test failed last night
#284
Retrofit job attributes to PMI v1.1
#297
make -j N support
#344
MPI_IN_PLACE bug in Allgatherv in MPE's collchk
#345
Re: MPI_IN_PLACE bug in Allgatherv in MPE's collchk
#346
execution time!!!!!
#349
Problem is partially solved
#378
ROMIO cross-compile bug on BGP
#396
producing logfiles for Jumpshot
#431
Re: [mpich2-dev] Hvector with Zero Blocks Asserts
#480
1.1b1 on icrunch
#494
Re: About the Jumpshot
#513
Is MPICH2 connection failure aware ?
#530
Re: run mpich on two architectures
#612
mpdboot trouble,pls help me,thanks a lot
#685
test/mpi/perf/commcreatep fails with Hydra
#754
very slow build on Mac OS-X
#788
aclocal_mpi.m4 missing macros
#846
problem parallelizing on nehalems
#867
Installing multiple versions of MPICH2 on a windows system
#890
mpich2-1.1.1p1 lock in MPI_Init_thread/MPI_Comm_spawn_multiple
#1009
POLA violation
#1020
problems with hydra/fork on freebsd
#1032
Does MPI_Recv use polling mechanism to receive message?
#1057
Bogus error message from mpdboot
#1062
iprobe problem
#1097
Remove hack in hydra configure to disable hwloc on Cygwin
#1103
Some test results on busy wait
#1108
mpiexec.hydra fails with totalview
#1446
MPI compiler wrappers failed with "make install prefix=PATH".
#1458
Hydra: race in killing processes
#1462
libmpi.a zero byte for i386 distribution
#1471
configuring with --with-mpi fails when using builddir
#1504
MPI_Comm_spawn, running out of resources with hydra on local machine
#1558
build failure in hwloc
#1600
Collectives fail when nemesis shared memory turned off on Mac
#1611
sending greater than 2^31 bytes
#1713
missing CXX types
#1717
missing function MPI_Get_library_version
#1811
Hydra: sessions not correctly tracked in pbs environments
#1884
Completion counter optimization
#1931
f90/datatype/get_elem_d failing with the PGI compiler
#1939
incorrect "-bind-to socket" behavior
#1944
PAMId: made the changes suggested by the review meeting
#1948
PAMId support of matched probe/recv functions
#1974
Using MPI_Buffer causes crash
#1990
error checking for neighborhood collectives
#1997
a dead lock occurs in nb_test.c
#2006
isendalive fails with --enable-g
#2029
NAG + NoCXX failures
#2052
MPI_Allgatherv does not detect aliased memory
#2060
strict builds fail to configure on Solaris
#2082
f77 spawn failures with NAG compilers
#2085
fortran collective tests hang on bgq using "per object" lock configuration
#2138
Add same_disp_unit info key for RMA window creation
#2154
Intel testsuite failures (Fortran)
#2158
multi_isendalive test failing
#2179
tests not building with --disable-shared
#2181
MPI_INTERCOMM_CREATE doesn't work with non-connected processes
#2219
Incorrect lower bound from MPI_Type_create_struct
#2243
comm_idup fails with mpich-portals4
#2292
Build failure with gcc-5.2 on Mac OS/X

Just like the query: wiki links, the parameter of this macro expects a query string formatted according to the rules of the simple ticket query language. This also allows displaying the link and description of a single ticket:

[[TicketQuery(id=123)]]

This is displayed as:

#123
MPIDU_Yield and MPID_Thread_yield

A more compact representation without the ticket summaries is also available:

[[TicketQuery(version=0.6|0.7&resolution=duplicate, compact)]]

This is displayed as:

#35, #44, #112, #125, #172, #173, #284, #297, #344, #345, #346, #349, #378, #396, #431, #480, #494, #513, #530, #612, #685, #754, #788, #846, #867, #890, #1009, #1020, #1032, #1057, #1062, #1097, #1103, #1108, #1446, #1458, #1462, #1471, #1504, #1558, #1600, #1611, #1713, #1717, #1811, #1884, #1931, #1939, #1944, #1948, #1974, #1990, #1997, #2006, #2029, #2052, #2060, #2082, #2085, #2138, #2154, #2158, #2179, #2181, #2219, #2243, #2292

Finally, if you wish to receive only the number of defects that match the query, use the count parameter.

[[TicketQuery(version=0.6|0.7&resolution=duplicate, count)]]

This is displayed as:

67

Customizing the table format

You can also customize the columns displayed in the table format (format=table) by using col=<field> - you can specify multiple fields and what order they are displayed by placing pipes (|) between the columns like below:

[[TicketQuery(max=3,status=closed,order=id,desc=1,format=table,col=resolution|summary|owner|reporter)]]

This is displayed as:

Results (1 - 3 of 1547)

1 2 3 4 5 6 7 8 9 10 11
Ticket Resolution Summary Owner Reporter
#2353 fixed type reference count issue balaji robl
#2351 fixed Missing CVAR category NODEMAP Yanfei Guo <yguo@…> gropp
#2350 fixed Segfault during request initialization aamer aamer
1 2 3 4 5 6 7 8 9 10 11

Full rows

In table format you can also have full rows by using rows=<field> like below:

[[TicketQuery(max=3,status=closed,order=id,desc=1,format=table,col=resolution|summary|owner|reporter,rows=description)]]

This is displayed as:

Results (1 - 3 of 1547)

1 2 3 4 5 6 7 8 9 10 11
Ticket Resolution Summary Owner Reporter
#2353 fixed type reference count issue balaji robl
Description

Edited from a few emails:

Cray thinks they found a bug in MPICH related to MPI type handling and eager sends. The code is creating a type, posting an Isend using that type, and then frees the type and response. The following text has been taken from the Cray emails:

The test case (from Comment #39) has clarified a few things for us:

1. This issue is also reproducible with the latest Vanilla ANL MPICH (3.2-Stable)
release. I tried building their trunk verson, but noticed a build issue (MPL
symbols..). So, yet to try with  the trunk. The failure signature (assertion in the
datatype segment code) with Vanilla ANL MPICH is consistent with the error
reported in this bug.

2. The issue is indeed random. It is a function of number of pending isends,
eager/rndv thresholds and whether the communication is on-node or off-node.
The odds of the failure increase when the test is doing on-node communication
with a handful of pending isends (~64) with small messages.

3. The failure is not observed with large data -- or when the library uses the rndv
protocol instead of the eager protocol.

Looking under the hood, one would expect the library to increment thedatatype object's ref-count for each pending isend operation. The ref-counts appear to be incremented correctly for each pending on-node, large message, isend operation. But, this does not seem to be the case for small message on-node transfers. This is a possible explanation for the random assertion failures and this is also true with the ANL MPICH code.

#2351 fixed Missing CVAR category NODEMAP Yanfei Guo <yguo@…> gropp
Description

The script that extracts CVAR information reports:

Extracting control variables (cvar) ... Category 'NODEMAP' referenced by 'MPIR_CVAR_NOLOCAL' was not found at ./maint/extractcvars line 110. Category 'NODEMAP' referenced by 'MPIR_CVAR_ODD_EVEN_CLIQUES' was not found at ./maint/extractcvars line 110.

There is no definition of the NODEMAP category in src/util/nodemap.

#2350 fixed Segfault during request initialization aamer aamer
Description

There are cases where MPID_Request_init segfaults on Mac OS with clang and also with a PathScale? compiler. The root of the problem is still unknown.

The attached C++ Bcast toy program reproduces the problem (thanks to Mark Davis <markdavisinboston@…> for the input). Running, for example, mpiexec -n 6 bcast_test_toy_segv 400 70 reproduces the problem.

1 2 3 4 5 6 7 8 9 10 11

Query Language

query: TracLinks and the [[TicketQuery]] macro both use a mini “query language” for specifying query filters. Basically, the filters are separated by ampersands (&). Each filter then consists of the ticket field name, an operator, and one or more values. More than one value are separated by a pipe (|), meaning that the filter matches any of the values. To include a literal & or | in a value, escape the character with a backslash (\).

The available operators are:

= the field content exactly matches one of the values
~= the field content contains one or more of the values
^= the field content starts with one of the values
$= the field content ends with one of the values

All of these operators can also be negated:

!= the field content matches none of the values
!~= the field content does not contain any of the values
!^= the field content does not start with any of the values
!$= the field content does not end with any of the values

The date fields created and modified can be constrained by using the = operator and specifying a value containing two dates separated by two dots (..). Either end of the date range can be left empty, meaning that the corresponding end of the range is open. The date parser understands a few natural date specifications like "3 weeks ago", "last month" and "now", as well as Bugzilla-style date specifications like "1d", "2w", "3m" or "4y" for 1 day, 2 weeks, 3 months and 4 years, respectively. Spaces in date specifications can be left out to avoid having to quote the query string.

created=2007-01-01..2008-01-01 query tickets created in 2007
created=lastmonth..thismonth query tickets created during the previous month
modified=1weekago.. query tickets that have been modified in the last week
modified=..30daysago query tickets that have been inactive for the last 30 days

See also: TracTickets, TracReports, TracGuide

Last modified 5 years ago Last modified on 01/23/13 17:10:34