Friday, March 24, 2006

xinfix

Trapped a possible write past end of string in xinfix which occurs if the new text input by the user is longer than the input string for the routine. I think this is what caused the xspec v11 crash under Scientific Linux 4.2 reported by Glenn Morris.

Notice that there is a more general problem here with v11 in that the string containing the command input by the user which is passed down to the xspec subroutines appears to have the
length of the command string. So, any replacement prompted from the user for part of this string has to be shorter than the original or problems will occur.

Note also that I need to remember that xspec gets the xparse routines from libxantcl not libxanlib - I wasted several hours trying to work out why changes placed in libxanlib weren't showing up when running xspec.

Keywords: xspec, heasoft

No comments: