L TES 01/27/04   1003
Test data required for models with errors - untested code.

L ENH 01/27/04   1004
Model errors need to be propagated through Plot object.

L ENH 01/27/04   1006
Code model.dat file for Switch and Scale parameters

L BUG 01/27/04   1007
Spacing in parameter link expression - fails to parse correctly.
(Also in xs11).

L UNI 01/27/04   1010
Limits to parameter numbers for table models.  Code will be very slow
or crash if many table model params are present.  Actual limit is about 30,
practical limit should be about 10.

L BUG 01/27/04   1013
Plot contour SEGV with certain inputs.  Fault is with setting contour levels.

L UNI 01/27/04   1021
Store command

L UNI 01/27/04   1022
Restore command

L BUG 01/27/04   1024
Log files currently not executable, '#' marks produced on wrong lines,
and sometimes not produced where needed.

L BUG 01/27/04   1025
Renorm doesn't seem to work correctly when convolution models are present,
check results against xspec11, confer with KAA. 

L UNI 01/27/04   1026
Help modes for local model and script directory.

L BUG 01/27/04   1028
Reinitialize mixing model after data are changed. 
Running mixing model, changing data, rerunning, and changing data back
may not give the same results in both cases.

L ENH 01/27/04   1031
ASCII data format: ability to read and fit ASCII data files.  Check with 
KAA what format flux2xspec accepts.

L BUG 01/27/04   1035
Error codes from Eqwidth may be incorrect when some of the spectra produce
results and others do not.

L UNI 01/27/04   1036
Thleqw unimplemented.

L ENH 01/27/04   1037
UserDummyRsp doesn't fill in various Response members that perhaps it 
could, such as efficiency.  This prevents peakrsid from working with it.

L UNI 01/27/04   1039
In xsShow, show rates only shows total, not net.

L BUG 02/02/04   1041
Eqwidth is not resetting defaults from previous user entered commands.

L BUG 03/02/04   1049
Project mixmodel seg faults on Linux only, when the second of 2 loaded
spectra has all its channels ignored.  This does not happen if a 3rd 
spectrum with noticed channels is loaded. Perhaps some sort of check for 
this situation should be done before mixing is attempted.

L BUG 05/13/04   1082
Steppar doesn't retain the previous values entered.

L BUG 06/24/04   1100
When many (~100) spectra loaded all in separate plot groups, simply calling
iplot may cause a seg fault.

L BUG 07/09/04   1110
Getting "Fit not current" warning when changing parameter of inactive 
model, but plotting data with active model. (Actually the only 
problem now is that changing an inactive model causes a "valid fit
does not exist" message to show up - 12/06)

M BUG 07/22/04   1119
Need to check "response none" handling for case of multiple sources.

L BUG 08/27/04   1137
X-range of plot data/model doesn't take into account if channels are
ignored.  ie --if chans 1-10 are bad and ignored, range still begins at
1.  This is not what xs11 seems to do.  (Need to consult KA)

L ENH 09/14/04   1142
Allow users to enter some special symbol for filenames in order to specify
relative paths from data rather than run directory.

M BUG 10/22/04   1153
Seg-fault occurs for an RMF that has all zeros in a row, and incorrectly
has f_chan and n_chan set to access them anyway.  This occurs through
divide by zero during normalizing.

M BUG 11/03/04   1158
For cases where spectra have been loaded in a complex way (see 1157),
show data command doesn't necessarily print output in order of increasing
spectrum number.  This should be easy to do now that bundledSpectra
array exists in DataContainer.

L BUG 11/23/04   1166
When dealing with named models, xsError command should output the name.
Currently it only outputs the par number.

M BUG 12/01/04   1169
During ModelContainer::Update, models shouldn't need to be recalculated
during stage where old responses have been removed and before new
responses have been attached.  This is related to issue of sending
spectrum number = 0 to model functions when dummy resp is attached.

L BUG 12/07/04   1172
Possible bug: need to check that LevMarq exits fit when all parameters
are pegged.  It should not prompt user to continue fitting.

L ENH 02/01/05   1186
Add ability to specify parameter limits as a function of other 
parameters, ie. param3 <= param2.

M BUG 03/22/05   1204
ignore/notice bug, doesn't accept " :10-20", where it should be using
previously entered spectrum range.

M BUG 04/06/05   1219
If users delete their src tree, or simply move the installation away
from it, they will run into problems building local models.  This results
from our use of tree structures in include statements.  One temp fix
is to manually hack in the path to the src tree into the HD_SRC 
variable in Makefile created by initpackage.

L BUG 04/18/05   1226
If spectra and backfile have quality cols with all 1s, it seg-faults during
groupArrays of background file.

M BUG 05/05/05   1246
Cannot not use arf command to specify arfs unless rmf is in place.  v11
however does allow this.

M BUG 05/06/05   1247
v12 currently bins this situation into 2 bins:
Q 0, 1, 0, 0
G 1, 1, -1, 1
it should be binning into 3 as does v11.

M BUG 05/09/05   1249
Certain SPI ARFs are causing seg-fault upon loading with multiple model
files (Linux only).  I've narrowed this down to ARFS corresponding to
sources 1 and 4 in C.Shrader's test data. (Turns out failure occurred
for ARFs having EXTVER = 2 in their headers.  This needs to be addressed
in CCfits.)

L UNI 05/12/05   1257
SPI data fakeit appears to be unimplemented for case of using dummy
response on fake data based on original data.

M BUG 05/16/05   1264
When fitting from a script, "Number of trials exceeded" message is
preceded by tcl jargon.

L ENH 05/24/05   1269
Allow default row specifier = 1 when dealing w/ type-II file with
only 1 row.

L ENH 06/09/05   1294
Perhaps allow user to load only part of response (via something like a
"pre-ignore" command) to save memory when using only a small part of the
matrix.  (Maurice Leutenegger)

L ENH 06/24/05   1298
Allow loading of type I data format with {n} notation, where n refers
to a spectrum extension number rather than a row.

M BUG 07/07/05   1308
When data group copies of model objects have their parameters untied,
they should follow the v11 method of reverting to state of dg1 pars.
In v12 their variable/frozen state reverts to model.dat setting.
(Phil Evans)

M ENH 08/01/05   1321
Martin Still's suggestion: add plot option to work in SI units
(axes in mJy and Hz).

L ENH 08/03/05   1324
Implement a new 'alias' command for naming table models.

L ENH 08/11/05   1327
Jeremy Sanders suggests improving the method to get the seed for the random
number generator. Incorporate perhaps the pid of the process, the ip address
of the computer, the uptime of the computer, or /dev/random if available.

M BUG 09/29/05   1359
In editmod, when faced with an ambigous command as described in the
manual, it is not prompting user.  Instead it just grabs the first
component.

M BUG 10/26/05   1367
Needs to better handle the error given by scalar QUALITY or GROUPING 
column given with type-2 files.

M BUG 11/01/05   1373
If a table model is loaded, then the file is changed and reloaded, the 
table doesn't change.  User needs to exit xspec in order to pick up the
change.  (Eric Gotthelf)

L DES 12/21/05   1392
Currently have both Fortran (ldaped.f, rdaped.f) and C++ (EmissionLines/
Apec.cxx) code that reads APEC files.  These should be unified at some
point.

M ENH 01/12/06   1401
Make acroread command selection more user friendly.  Ideally, Xspec.init
would be customized on installation, but perhaps an output message to
user upon acroread failure would suffice.

M BUG 01/20/06   1406
v11 allows this type of syntax inside a script when running it from
"source" command: "fakeit 1 & / & / & /", while v12 requires the use
of the "@" symbol.  I assume this is true for other multi-line commands
as well.  It appears to be due to XSparse::executingScript not being
set to "true" when run from "source" command.  (Simon Vaughan)

M BUG 01/27/06   1410
Tclout plot contour is not working.  Plot::saveArray function needs
to add a special case for contour plots.

M BUG 01/30/06   1413
After untie on a linked parameter, the parameter should keep all its
values the same (except of course delta fit) and not reinitialize them
from model.dat file. (Jeremy Sanders)  (Note: this repeats issue 1308)

M ENH 02/21/06   1431
Possibly add f90 capability to initpackage.  See P. Leopardi's
02/17 email via KAA.  This will require conforming changes in 
Xspec/BUILD_DIR/Makefile-std as well as some possible hmakerc changes.

M BUG 03/02/06   1435
initpackage recompiles any model source files, even if the object
files are more recent.

M BUG 03/13/06   1438
When a model is applied to 2 or more spectra whose responses have 
different input energies, "plot model" only seems to plot one of them.
The trouble is in the Model version of Plot::initializePlotGroup.

M ENH 03/17/06   1439
Expand eqwidth to allow for calculations of multiplicative 
components.  In this case the continuum is defined as all additive
and multiplicative components that are multiplied by the particular
multiplicative component under study.

M ENH 03/23/06   1443
Add a more flexible format for the energy ascii file.  Lines could also
contain [<number of bins>] [log|lin] following the energy value.  
Number of bins would apply to the range between the preceding energy
value and this.

M BUG 04/19/06   1449
Undo is seg-faulting when applied after the removal of a type-II 
spectrum from a still existing type-II data set.

M UNI 04/24/06   1451
The weight command needs to take an additional parameter for spectra
range.  This ability was added into v11 at some point.  Also see 1454. 
(Bin-Bin Zhang)

M ENH 04/27/06   1456
Possibly add and maintain specialized tcl variables (ie. $xs_filenames, 
$xs_datasets, etc.) for particular data so as not to require user to
go through tclout and $xspec_tclout for everything.

M BUG 05/11/06   1466
Need to look into undo command behavior when autosave frequency is set
to something other than 1.

M BUG 09/11/06   1495
When running system commands (ie. ls, date) from XSPEC's shell, they get 
echoed with full path at the bottom of the output. (Roderick Johnstone)

M BUG 09/13/06   1497
Several save command issues: remove default insertions of "cd <fullpath>"
in save output (see KAA 9/12 email).  Write separate ignore commands
on separate lines.  Sort model by source number.  (Steve Snowden)

M BUG 10/25/06   1514
Undo is not working on a "data none" when there was more than 1 detector
assigned to the erased spectrum.

M BUG 10/26/06   1516
The chatter level of the "pegged parameter" message should be adjusted
to allow it to be killed off without removing the regular output.

M BUG 12/11/06   1534
Newpar seg-faults if it tries to create a parameter link that would
result in a circular dependency.

M ENH 12/18/06   1537
Implement a new StatMethod class for Kolmogorov-Smirnov statistics.
This should be a template class to allow for variations.

M BUG 01/02/07   1543
Addcomp and editmod should not allow a mix component to be added to an
inactive model.  The best place to check for this appears to be
Model::insertComponent.

M ENH 01/18/07   1550
Add a setplot option to put each dataset (and model) into its own window.
Thus if there were two datasets and the user did a plot data there would 
be two windows (stacked vertically).

M BUG 03/06/07   1565
Currently QUALITY columns in background files are ignored.  Instead,
an overall quality should be equal to the worst of either the source
or background.

M ENH 05/09/07   1584
Provide a shortcut to the response syntax when reusing the same
response.  ie. response 2:1 = 1:1   (Mark Bautz)

M ENH 05/25/07   1591
Add ability to modify current covariance matrix by a constant multiplier.

M BUG 05/30/07   1594
If a data file has a "NULL" entry, this is read in as a NaN and seems
to be independent of whether or not the address of a nullVal is sent
to the CCfits read function (Jeremy Sanders).

M ENH 05/31/07   1596
As an example, for case of model A(B+C) followed by editmod A(B)+C,
it'd be better if C could retain its parameter values.  As it is, 
they are being reset. (J. Sanders)

M BUG 06/19/07   1609
In CCfits, Column::readArrays functions should be able to handle scalar
column case as well as vector columns.

M BUG 06/28/07   1614
With latest Mozilla and Firefox, and new window or tab is now opened
for each help command whether or not a browser window is already
open.

M ENH 07/13/07   1618
Convert plot margin to show a cumulative distribution.  For 2-D case,
the proper method for this still needs to be determined.

M ENH 08/31/07   1639
Add a pi zero decay model for use with data from GLAST. (Hakki Ogelman)

M BUG 09/12/07   1640
When 2 parameters are linked and they have different hard limits, the
tighter of the 2 ranges should always take precedence.  Otherwise this
can particularly cause problems for table models, producing an out
of range error. 

M ENH 10/10/07   1654
For plot efficiency, have it take into account ignored channels the way
v11 does.  (Goro Sato)

M ENH 10/16/07   1657
Add OpenMP pragmas in key places to allow multithreading.

M ENH 11/16/07   1669
Add ability to assign mix models to source numbers other than 1.  This
will require passing source number info to the mix functions, and having
ModelContainer deal with more than 1 copy of a given mix model.

M BUG 11/29/07   1673
When using error command on periodic parameters, bracketting should
take smaller (perhaps linear) steps.

M ENH 12/03/07   1675
Steppar should save the parameter values if a new minimum is found, 
and at the end the user should be prompted whether to replace the 
parameters with the new minimum.  Also, tclout steppar needs an option
delstat to give the delta statistic column of the steppar output.

M ENH 12/03/07   1676
Add new chain/fit hybrid capability.  Something like:
   chain parameters 1 3 4
   chain run chain.fits
where the chain would be produced for 1,3,4, and all other pars would
be fit at each step.  (See 12/03 emails)

M ENH 01/14/08   1685
Allow convolution components to operate directly on multiplicative 
components, eg. (CM)A in addition to the currently allowed C(MA).
(See 1/13 email for details)

M BUG 02/04/08   1690
The tcl flush command does not seem to be available from inside a v12
shell, nor does it show up among the output list of "info commands".
(Dave Strickland)

M ENH 02/25/08   1699
Add the html help tarball to the distributed version of XSPEC, and have
it install itself in the spectral/help dir.

M BUG 03/18/08   1706
When new version of CFITSIO is checked in, the definitions for 
LONGLONG_MIN and MAX will be moved from fitsio2.h to fitsio.h.  
CCfits will need to make conforming change to the include headers in
Column.cxx

M DOC 03/27/08   1711
For the model command, the options active/inactive/none can now be
applied to the unnamed model by entering "unnamed". 

M BUG 03/27/08   1712
Data command bug: should be allowed to delete multiple spectra (and
not just highest numbered ones) by doing something like "data
none none/".

M BUG 04/25/08   1716
Some seg-faults unconvered with new exhaustive models library test
script: kerrd on SL4.0 with data set sis1, absori and nsa on 64-bit
Linux and IntelMac/leopard develop builds.

M ENH 04/29/08   1717
Add feature to have autocompletion of Tcl commands and variable names.

M BUG 04/29/08   1718
The "help minuit" option does not work if help is set to access online
files rather than local PDF.

M BUG 05/23/08   1721
When an error causes the error command to exit early, it should set the
particular error bound to 0.0 rather than leaving it untouched.
Otherwise it will misleading report the value of an earlier error 
command run. (Stefan Czesla & Christian Schneider)

M ENH 06/03/08 1722
Allow setplot group to group spectra which have the same energy bins but
different channels of bad quality. (Carlos Gabriel)

M ENH 06/09/08  1724
Allow the following syntax for assigning a source N to a model:
model N = <name>
This is to allow multiple sources to use the same model without having
to link parameters.

M DOC 06/10/08  1725
Add new option "tclout goodness" to manual.

M UNI 05/09/08   1728
Convolve operator remains unimplemented in mdefine (see 1603 for the
rest of mdefine entry).

M BUG 05/02/08   1730
The incomplete Gamma function algorithm needs to be replaced by a more
robust version (similar to what the Gnu libgsl has).  Otherwise
parameters can drift outside the range of the approximation's validity.  
This was observed to cause distortions in the cutoff powerlaw model.
(Joern Wilms, Moritz Boeck, and Manfred Hanke)

M UNI 07/07/08  1733
Add extended file name syntax handling to CCfits.  (See Gudlaugur 
Johannesson's fix for row filter handling as a possible implementation.)

M DOC 07/07/08  1734
The xspec usage of "angular size distance" is ambiguous.  It can mean
"proper motion distance" whereas xspec intends it to mean "angular 
diameter distance".  (Randall Smith)