Recompilation instructions. Most patches require recompilation of XSPEC. Please follow the instructions below if necessary

1. If you have compiled xspec for other machine architectures since you originally installed it, you will need to reconfigure the package for your current machine. To do this:
   • cd {distribution root}/src/BUILD_DIR
   • ./hmake distclean
   • ./configure

2. Set environment variables for development (here are the csh & variant instructions):
   • setenv LHEASOFT {distribution root}/{architecture}, where architecture is the architecture specific directory, such as Linux_2.2_i686 or SunOS_5.6_sparc
   • source $LHEASOFT/lhea-init.csh

3. If you build with local models make sure to setenv LMODDIR to your local model directory.
4. Change directory to {distribution root}/src/spectral/xspec/src

5. type "hmake", followed by "hmake install". If you cleaned the distribution, then this will recompile the xspec binary from the beginning (but not the libraries upon which it depends).

Using addcmp to add a convolution model ahead of a multiplicative model produces a spurious error message. Our thanks to Ian George for pointing this out. Bug report added on Jun 7, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/MODEL_COMP/addcmp.f and rebuild xspec.

Reading a spectrum with some channels having bad quality and with grouping defined (using eg grppha) can sometimes produce a memory allocation warning and subsequent seg fault. Our thanks to Mariano Mendez for reporting this bug. Report added on Jun 8, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/FILE/rdgrpc.f and rebuild xspec.

Any model systematic errors set using the systematic command are not saved by the save command. Our thanks to Miriam Krauss for reporting this bug. Report added on Jun 10, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/COMMAND/autpro.f and $LHEASOFT/../src/spectral/xspec/src/COMMAND/wrtpro.f then rebuild xspec.

Doing a fakeit with no background file can cause a floating overflow. Report added on Jul 1, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/DATA/calfak.f then rebuild xspec.

Reading a spectrum file with an ANCRFILE keyword longer than 100 characters causes a crash under OSF1. Our thanks to Glenn Morris for reporting this. Report added on Aug 3, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/FILE/gtarin.f then rebuild xspec.

The xmmpsf model doesn't behave correctly if new datasets are read or channels ignored/noticed after the model has been defined. Our thanks to Glenn Morris for reporting this. Report added on Aug 19, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/functions/xmmc.f then rebuild xspec.

/* or none don't work correctly when prompted for a replacement filename (eg in the data command). Our thanks to Ed Kellogg for reporting this. Report added on Sep 5, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/COMMAND/cutils.c then rebuild xspec.

tclout plot doesn't write numbers to the full available accuracy. Our thanks to Jeremy Sanders for reporting this. Report added on Sep 5, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/PLOT/xpltit.f then rebuild xspec.

Reading a spectrum with some channels having bad quality and with grouping defined (using eg grppha) produces a memory allocation warning when a second file is read. Our thanks to Fill Humphrey and Mariano Mendez for independently reporting this bug. Report added on Sep 10, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/FILE/rdgrpc.f and rebuild xspec. Note that this supersedes 11.3.1b.

In the help for the abund command the abundance of Si is given as 3.35e-5 for the grsa option. The actual value used in the code (and which is correct) is 3.55e-5. Our thanks to Stefano Ettori for spotting this typo. Report added on Sep 17, 2004.

The following is an enhancement not a bug fix. All the models that use the mekal code (eg mkcflow) can be made to use the APEC code by setting the switch parameter to 2. Report added on Oct 11, 2004.

The enhancement is to replace $LHEASOFT/../src/spectral/xspec/src/functions/ldaped.f, $LHEASOFT/../src/spectral/xspec/src/functions/sumdem.f, $LHEASOFT/../src/spectral/xspec/src/functions/smdem2.f, $LHEASOFT/../src/spectral/xspec/src/functions/xsvape.f and rebuild xspec.

If a non-xspec (ie system-level) command is given after fakeit in a script that command will be ignored. The work-around is to include an xspec command (eg show) between fakeit and the non-xspec command. Our thanks to Davide Donato for bringing this to our attention. Report added on Dec 2, 2004.

The pileup model core dumps with a floating exception under some circumstances. Our thanks to David Hui for pointing this out. Report added on Dec 3, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/functions/pileup.f and rebuild xspec.

If the abund filename option has been used then save model (and auto-saving) does not write out the filename. Our thanks to Fill Humphrey for spotting this. Report added on Dec 6, 2004.

The fix is to get the latest patch tarfile and rebuild xspec.

There is no way of saving the results of the goodness command and writing it out. This has been fixed by adding a tclout goodness option. Our thanks to Fill Humphrey for suggesting this. Report added on Dec 6, 2004.

The fix is to get the latest patch tarfile and rebuild xspec.

If all parameters peg then the user is asked whether to continue fitting. However the fit can go nowhere from this point so this question only results in an infinite loop if the user has set query yes. Changed to end the fit in this case. Our thanks to Fill Humphrey for pointing this out. Report added on Dec 6, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/FIT/lvmrun.f and rebuild xspec.

The compbb model gives incorrect results if the response energy range is changed but the kT and kTe parameters are not. This can happen if spectra from different instruments are being fit simultaneously. Report added on Dec 17, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/functions/compbb.f and rebuild xspec.

When running fakeit based on a data file with grouping the resulting spectrum can have some very large error bars if this is the first fakeit of an xspec session. Our thanks to David Buote for reporting this. Report added on Dec 21, 2004.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/DATA/wrtfak.f and rebuild xspec.

The STAT_ERR column is incorrect in faked background files which have Gaussian, not Poisson, errors and where the BACKSCALE keywords in the source and background files differ. This situation probably never occurs. Report added on Dec 30, 2004. UPDATED on Jan 5, 2005 to fix typo.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/DATA/calbfk.f and rebuild xspec.

Performing a fakeit none before any datasets have been read in can occasionally generate a seg fault and crash xspec. Our thanks to Jeremy Sanders for tracking this down. Report added on Jan 18, 2005.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/DATA/fakeit.c and rebuild xspec.

The label giving the minimum and levels is not written out by plot contour. Report added on Feb 8, 2005.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/PLOT/xpltit.f and rebuild xspec.

xhistory doesn't write out the fit package. Our thanks to Robert Radocinski for pointing this out. Report added on Feb 8, 2005. Updated on Feb 9, 2005 to fix bug that sigma values were not written to the fit package.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/FIT/lvmfit.f, $LHEASOFT/../src/spectral/xspec/src/FIT/lvmrun.f and rebuild xspec.

The error command is inefficient when the parameter at issue has hard and soft limits that differ. Our thanks to Pat Broos for reporting this. Report added on Feb 24, 2005.

The fix is to replace $LHEASOFT/../src/spectral/xspec/src/FIT/clcunc.f and rebuild xspec.