next up [*] [*]
Up: Revision History for Version Previous: XSPEC v11.1 issues fixed

XSPEC v11 issues fixed in 11.1

XSPEC does not build with gcc owing to an internal compiler error

Xspec will SEGV and dump core if parameter linking expressions are entered in Reverse Polish Notation rather than Algebraic Notation

The Xanadu/HEAsoft help system may cause a SEGV if help is entered twice during a session.

(c) The fakeit command ignores auxiliary response files (d) The fakeit command does not process correctly within a tcl script.

Methods for using xspec interactive commands such as model, fakeit, and newpar inside tcl procedures are not documented.

XSPEC will SEGV if an attempt is made to invoke a script using the syntax within another script invoked with the syntax.

When a negative multiplicative factor is used to link parameters, if the model is saved, the save file will read the factor back as an offset, not a multiplicative factor.

Invalid filename entered for table model causes SEGV

XSPEC does not recognize correct OGIP / 1992a files (as indicated by the PHAVERSN keyword) if the file is lacking the HDUCLASS keyword

Spurious scaling factors are generated when reading models back from``save" files.

XSPEC will SEGV if the character is typed in on a line by itself.

XSPEC goes into infinite loop if an incorrect plotting device is entered. Noted: 2000 April 6.

The URL on the PGPLOT "splash" screen for the bugs page is out of date.

The entry prompt for response files in fakeit ignores files given by absolute paths.

tclout error sets $xspec_tclout to garbage values.

tclout plot option does not process correctly.

Xspec crashes (SEGV) after processing a tcl loop within a command (.xcm) file

The meka model shows no emission lines as it uses an input file which recent versions of cfitsio cannot read. This model is in any case obsolete and should only be used for historical comparisons.

The cemekl and cevmkl models do not integrate over a range of temperatures as claimed.

XSPEC crashes (SEGV) intermittently in the dialogue that prompts for replacement filenames.

XSPEC crashes (SEGV) when saved model expressions of sufficient length are read back into XSPEC from an .xcm script

For fake datasets, count rate was output as integer if no counting statistics option selected.

(1) XSPEC crashes (SEGV) when attempting to load an .xcm script from within an .xcm script using either the @ syntax or from the command line. (2) XSPEC crashes (SEGV) when saved model expressions of sufficient length are read back into XSPEC from an .xcm script

(Model string "jumbo patch") XSPEC underallocates memory for the strings describing extremely complex models.

The apec and vapec models contain an error that leads to subtly incorrect spectra.

The F value is incorrect if more than one parameter has been added.

The model string printed by the show command is not properly re-initialized when the model is changed.

XSPEC crashes (SEGV) if user requests a fake dataset without us. If the user supplies insufficient arguments to fakeit from a script, XSPEC will now exit with an error (prior to this it would SEGV).

XSPEC will not perform fits if all channels are ignored in any dataset loaded.

XSPEC will crash if a model is defined and the last dataset read has no response matrix.

The tclout rate option did not work correctly.

The MKCFLOW model causes XSPEC to crash if the lower temperature limit is less than about 0.02 keV.

(1) Commands of the form ignore 5.0-** ignore channels with wavelengths less than, rather than greater than the specified value if setplot wave has been given. (2) If multiple datasets are in use, then even-numbered datasets will interpret the range as energy instead of wavelength.

steppar does not work correctly if stepping is performed over more than three parameters.

In the APED model, the helium abundance is varied with the other abundances instead of remaining constant at the Solar estimated value.

(1) Files with complex patterns of "bad" channels may cause xspec to crash when the ignore command is invoked. (2) Version string updated to print patch level of running xspec version.

EBOUNDS arrays are constructed incorrectly for datasets with grouped channels if the arrays are decreasing with channel number.

Fixes for fake file generation. (a) XSPEC crashes (SEGV) if more than about 100 files are simulated (e.g. from a large OGIP-Type II file). (b) simulating background causes corrupted memory if no dataset has been read in.

Fixes xspec tcl scripts implementing addline and modid commands.

In XSPEC v11 convolution components work on the model*(effective area). This was changed from v10 to enable pile-up to be implemented as a convolution model. However, this change can introduce features when using other types of convolution models. This patch reverts xspec to the v.10 behaviour.

XSPEC crashes intermittently in the command tclout model.

(1) XSPEC crashes on fakeit command if no datasets are loaded. (2) Verbose output printed to log file even when chatter level is set to 0 (3) XSPEC crashes in the replacement file dialogue for new table model if user enters a blank line. (4) some linux version builds fail to compile the xanlib library because of the lack of some trigonometric fortran functions.

XSPEC crashes on DEC/Compaq platform if spectra with NULL or INDEF values are present in the COUNT or RATE column, as occurs with XMM/RGS files.

The grad model contains several bugs, causing the mass obtained from fitting the model to the observation to be over-estimated by a factor 1.4. These bugs have been fixed and a new parameter (par6) added to make clear the distinction between the old and new models. The online help has been amended accordingly.

When operating in wavelengths (after setplot wave) XSPEC does not ignore correctly if and attempt is made to specify the ignore range in wavelength (eg ignore 5.0-6.0) and the input dataset is stored in increasing wavelength (e.g. XMM-Newton RGS spectra).

gain fit option fails in command scripts.

fakeit none causes XSPEC to halt if there are no datasets previously defined and the user enters an invalid file.

Fixed but not previously reported

Neither the ** option nor giving a wavelength outside the dataset range worked correctly when ignoring or noticing after setplot wave.

There was a spurious zero width energy bin produced by the extend high command.

The background scaling was not being taken into account when using the lstat statistic.

When using the annealing method the wrong parameters were updated if any had been frozen.

In steppar the best/current choice had to precede the log/nolog choice. Also, the write statement for the output for each step did not work correctly for more than 3 parameters.

The save command now did not save gain shift information when gain was not being fitted.

There were illegal blanks in the TFORM# keywords in the eigen*.fits for the NEI models.

In linefile.fits the TRANS column was A8 which failed to newer versions of cfitsio.

There was a minor typo in photo.f. It would only have made a difference if there was more than one edge in a single energy bin, which probably never happens.

There was an error in the table interplation algorithm which showed up if parameters had values less than about 1e-5.

An error was corrected in comptt that was introduced in July 99. This has been fixed along with the original problem that the earlier incorrect fix was supposed to handle.

Only the first mixing model was available.

There was a bug that gave an incorrect equivalent width for all datasets except the first if multiple datasets were in use and the lower end of the continuum range used corresponded to the lowest energy in the response.

The model cevmkl was actually using the C abundance as the density. This would only have mattered if a high density plasma was being analyzed.

next up [*] [*]
Up: Revision History for Version Previous: XSPEC v11.1 issues fixed
Ben Dorman