Compile with the following with '-c -standard-semantics'
integer, allocatable :: y(:)
y = merge(1, 0, [0.d0] > 0.d0)
Thank you for reporting this error and for the convenient reproducer. I confirmed the error and reported it to Development (internal tracking id below) and will keep the thread updated regarding a fix. The option -assume realloc_lhs implicit with -standard-semantics produces the error. If not dependant on that option or at least as a work-around to permit successful compilation of the affected source file, add -assume norealloc_lhs *after* -standard-semantics to override the implicit setting with -standard-semantics.
(Internal tracking id: DPD200237427)
(Resolution Update on 01/23/2014): This defect is fixed in the Intel® Fortran Composer XE 2013 SP1 Initial Release (2013.1.0.080- Linux)
Please pardon the delayed update. This defect was fixed in the Intel® Fortran Composer XE 2013 SP1 Initial Release (2013.1.0.080 - Linux)