next up [*] [*]
Next: XSPEC v11 issues fixed Up: Revision History for Version Previous: XSPEC v11.2 issues fixed

XSPEC v11.1 issues fixed in 11.2

On some Linux systems garbage characters are generated when fitting certain models (eg zpow, refsch). This is due to uninitialized status variables in a number of routines.

The file written by WENV after IPLOT CONT cannot be read back into QDP.

There is a problem reading BACKSCAL and AREASCAL columns from type II format spectral files.

There is a very rarely triggered bug in the error command which causes a crash.

Under Linux the data syntax filename{number} or filename{*} does not work.

XSPEC will crash if asked to re-open an autosave file in a directory to which the user does not have write permission.

Under Linux the pexrav model can cause the fitting process to fail (with warning messages from svdcmp).

The pileup model produces subtly wrong answers if there is pile-up beyond 2 photons.

The sedov model doesn't work under some Linux systems.

Under Linux several of the NEI models cause segmentation faults under some circumstances.

The default parameter limits on the mkcflow and vmcflow models are wrong and can cause XSPEC to crash. This is caused by a combination of incorrect parameter limits and coding errors.

Building XSPEC without local models and then setting the LMODDIR environmental variable to link in local models does not work.

If the dataset has vector areascal or backscal these are not treated correctly by the ignore command.

If the dataset has vector areascal or backscal and it is grouped then the final bin will have incorrect areascal and backscal values.

Fakeit fails for a file with vector values of more than two of quality, grouping, areascal, or backscal.

No warning is generated if an ARF has fewer energy ranges than the RMF with which it is paired. Technically this isn't a bug but it seems to be quite easy to create Chandra ARF/RMF pairs with this problem - the diagnostic is that the gain appears to be wrong.

fakeit does not allow its output file to be saved into a directory specified by an absolute path (i.e. beginning with a "/").

(user interface patch) XSPECs internal strings within the data command are not properly initialized. Additionally, buffer overflows can cause unexpected crashes on some platforms. Also, XSPEC scripts that refer to non-existent files do not properly prompt the user for a replacement filename.

XSPEC with MINUIT does not work under Redhat Linux 7.1.

The pileup model parameters do not vary when fitting.

The Anders and Grevesse Cl, Cr, and Co abundances are incorrect.

There is a bug in the notice command that is triggered if channels are noticed in a dataset and there are subsequent datasets which do not have channels noticed. eg if there are 3 datasets read in and the first has channels ignored then a notice 1:1-** will trigger the bug and corrupt the responses of the second and third datasets.

There are errors in the background spectrum contribution to the calculation of $\chi^2$ for the non-standard weighting options (gehrels, churazov, model).

Contrary to the impression given by the help, the Tuebingen absorption models do not use the ISM abundance ratios given in the Wilms et al. paper but use the standard ratios set by the abund command.

The help on the cosmo command is incorrect in the case that a non-zero cosmological constant is given. It is not true that q0 must be 0.5. Infact q0 is ignored and it is assumed that the universe is flat with $\Omega_{matter} = 1 - \Omega_\Lambda$. The workaround is to leave q0 = 0.5 and forget about it.

The nei models seg fault under Linux if the response energies extend outside the range of validity of the model.

The normalization of the redge model can depend on the size of the energy bins in use.

The mkcflow model varies the He abundance in parameter 3 along with the other elements. This is inconsistent with other collisional plasma models which fix He abundance to Solar (except for those models with all elemental abundances as free parameters).

XSPEC crashes under Linux if the same type II spectral file is read using the {} syntax to specify the spectrum more than once in an XSPEC session.

The pileup model produces subtly wrong results if the energy range in the response does not start at zero.

A subtle bug can arise when reading a new dataset into XSPEC to replace a current dataset. If both datasets have the same number of ungrouped channels and the same number of grouped bins then XSPEC will apply the grouping array from the original dataset to the new one. This may not be the correct behaviour. The ignore status will also be copied from the old file to the new one. This also may not be correct.

The error command can get stuck in an infinite loop.

next up [*] [*]
Next: XSPEC v11 issues fixed Up: Revision History for Version Previous: XSPEC v11.2 issues fixed
Ben Dorman