This commit makes two changes to improve the usability of the frontend in the open
environment:
1) It increases the maximum supported screen size. To deal with limits on stack
allocated memory (one occurrence in the code), the frontend limited the maximum
number of rows and columns to values appropriate for the hardware of the day.
While this change doesn't remove the limits, it increases them to values
appropriate for modern 4k graphics hardware.
2) It refines an executability pre-test of the desired backend (i.e, 'batchvision')
to allow the successful execution of $PATH resolved backends. As originally
written, this test was always performed; however, the 'access' routine used to
perform that test is not $PATH aware. As a result, all $PATH relative backend
selections were rejected unless a suitably named executable file could be found
in the current directory. The change implemented here limits the use of 'access'
to only those executables that appear to name a specific file (as evidenced by
the presence of '/'). This seems a reasonable compromise since the original
reason for the 'access' pre-test is lost in antiquity and code paths that bypass
that test already exist.
This commit makes two changes to improve the usability of the frontend in the open environment:
1) It increases the maximum supported screen size. To deal with limits on stack allocated memory (one occurrence in the code), the frontend limited the maximum number of rows and columns to values appropriate for the hardware of the day. While this change doesn't remove the limits, it increases them to values appropriate for modern 4k graphics hardware. 2) It refines an executability pre-test of the desired backend (i.e, 'batchvision') to allow the successful execution of $PATH resolved backends. As originally written, this test was always performed; however, the 'access' routine used to perform that test is not $PATH aware. As a result, all $PATH relative backend selections were rejected unless a suitably named executable file could be found in the current directory. The change implemented here limits the use of 'access' to only those executables that appear to name a specific file (as evidenced by the presence of '/'). This seems a reasonable compromise since the original reason for the 'access' pre-test is lost in antiquity and code paths that bypass that test already exist.