21.0 Showstoppers

Ben Wing wrote: I have access to Win95, 98 and NT4.0. I routinely test the installer on all of these. I do not have access to NT3.51, so I have no way to test. It seems pretty clear from the reports coming in that it simply won't work on 3.51. I will add an OS-version check to the installer and bail out if OS == NT && version < 4.0, unless somebody can tell me what needs to be done to make it work on WinNT. I have stuck the installer in my personal little bit of web-space, but have always hoped that somebody would pick it up and put it on the main XEmacs site. I submitted a patch that fixed a bunch of the Win95/98 subprocess brokenness, but it was rejected for (at least) two reasons: (A)it was a bit kludgey, and (B) there is some disagreement about the need for "cmdproxy" and the ability of the existing XEmacs subprocess code to spawn 16-bit subprocesses. The kludge was there because I didn't understand why the variable "win32_quote_process_args" didn't work as advertised, and I introduced another variable to do the same thing. The problem was "vars_of_ntproc" not getting called from emacs.c. On the other issue, I am fairly certain that "cmdproxy" is *not* necessary; "command.com" should do just fine. "ntproc.c" checks whether the process being spawned is "16-bit" (if it has the ".com" extension it is considered 16-bit) and if so uses cmdproxy to spawn it. However when I simply took this piece of code out I found that XEmacs can apparently spawn "command.com" just fine. It's perhaps a bit confusing to follow the code, the "spawnve" that is called at line 734 in "callproc.c" is *not* the Win32 spawnve function, but the spawnve function in "ntproc.c", which is capable of spawning a 16-bit subprocess. (Try it and see....) Anyhow, I'm glad you consider the broken subprocesses on 95/98 as a problem worth fixing. I will clean up my rejected patch to make it less kludgey and resubmit it; maybe it can still make it into 21.0. I'll leave these to somebody else!