Noticed while working on the stress tests. A unit test checks that 128 Mb startup parameter is picked up, but it seems that ~2GB is the max that R native accepts, even on a 64 bits windows box.
Comments: R_max_memory is in gnuwin32 folder, extra.c, malloc.c, system.c So, looks like INT_MAX default for R_MAX_MEMory is not modified on 64 bits (but it is down to ~1500Mb on 32 bits) cmdlineoptions needs to be called; only place where modified. But it is called only from RTerm and RGui for windows applications. How did StartupParameter ever work on Windows, then??? Other avenues are via the RStart struct, but max_vsize is never used to change R_max_memory. One wonders what this struct is for, if command line arguments are the only way to affect settings. Let's look at what is available in R.dll, exported R_common_command_line R_set_command_line_arguments
Comments: R_max_memory is in gnuwin32 folder, extra.c, malloc.c, system.c So, looks like INT_MAX default for R_MAX_MEMory is not modified on 64 bits (but it is down to ~1500Mb on 32 bits) cmdlineoptions needs to be called; only place where modified. But it is called only from RTerm and RGui for windows applications. How did StartupParameter ever work on Windows, then??? Other avenues are via the RStart struct, but max_vsize is never used to change R_max_memory. One wonders what this struct is for, if command line arguments are the only way to affect settings. Let's look at what is available in R.dll, exported R_common_command_line R_set_command_line_arguments