further to this, I'm definitely finding problems with the system(5) in
fxposix.ri
For a start, I don't understand why the mode argument is allowed to be a
fie. An integer, a string, or null makes sense, but a file? That's not the
problem...
You may remember I mentioned that system() seems to be losing the last
argument from the list. That's confirmed for the first call to system().
The next problem I've just noticed (on Windows) is spurious junk ending up
on the command line of the called process. However not the first call; the
2nd, 3rd etc call to system progressively passes junk across.
Here's a dump of caller and callee's ideas of the arguments, printed with
image() to stderr (and manually broken into shortish lines just for the
email)
CALLING "c:\\Progra~1\\WacGen\\tools-trunk\\bin.t\\wacro"
"-c"
"-I" "c:\\Progra~1\\WacGen\\tools-trunk/lib"
"-I" "c:\\Progra~1\\WacGen\\lib-trunk/include"
"-o" "wac0000.tmp"
"-f" "../genero.4gs/alien.src"
"FORMCOLORS=fourgen" "ROWID_TYPE=integer" "FOURGEN=1"
CALLED:
"-c"
"-I" "c:\\Progra~1\\WacGen\\tools-trunk/lib"
"-I" "c:\\Progra~1\\WacGen\\lib-trunk/include"
"-o" "wac0000.tmp"
"-f" "../genero.4gs/alien.src"
"FORMCOLORS=fourgen" "ROWID_TYPE=integer"
------
CALLING "c:\\Progra~1\\WacGen\\tools-trunk\\bin.t\\wacro"
"-c"
"-I" "c:\\Progra~1\\WacGen\\tools-trunk/lib"
"-I" "c:\\Progra~1\\WacGen\\lib-trunk/include"
"-o" "wac0001.tmp"
"-f" "../genero.4gs/brw_dump.src"
"FORMCOLORS=fourgen" "ROWID_TYPE=integer" "FOURGEN=1"
CALLED:
"-c"
"-I" "c:\\Progra~1\\WacGen\\tools-trunk/lib"
"-I" "c:\\Progra~1\\WacGen\\lib-trunk/include"
"-o" "wac0001.tmp"
"-f" "../genero.4gs/brw_dump.src"
"FORMCOLORS=fourgen" "ROWID_TYPE=integer"
"esources]\"/sand/devel/master/auth/FourJs/resources\"__DEBUG_INIFILE__"
------
CALLING "c:\\Progra~1\\WacGen\\tools-trunk\\bin.t\\wacro"
"-c"
"-I" "c:\\Progra~1\\WacGen\\tools-trunk/lib"
"-I" "c:\\Progra~1\\WacGen\\lib-trunk/include"
"-o" "wac0002.tmp"
"-f" "../genero.4gs/caller.src"
"FORMCOLORS=fourgen" "ROWID_TYPE=integer" "FOURGEN=1"
CALLED:
"-c"
"-I" "c:\\Progra~1\\WacGen\\tools-trunk/lib"
"-I" "c:\\Progra~1\\WacGen\\lib-trunk/include"
"-o" "wac0002.tmp"
"-f" "../genero.4gs/caller.src"
"FORMCOLORS=fourgen" "ROWID_TYPE=integer" "ac0001.tmp"
------
CALLING "c:\\Progra~1\\WacGen\\tools-trunk\\bin.t\\wacro"
"-c"
"-I" "c:\\Progra~1\\WacGen\\tools-trunk/lib"
"-I" "c:\\Progra~1\\WacGen\\lib-trunk/include"
"-o" "wac0003.tmp"
"-f" "../genero.4gs/chk_cncl.src"
"FORMCOLORS=fourgen" "ROWID_TYPE=integer" "FOURGEN=1"
CALLED:
"-c"
"-I" "c:\\Progra~1\\WacGen\\tools-trunk/lib"
"-I" "c:\\Progra~1\\WacGen\\lib-trunk/include"
"-o" "wac0003.tmp"
"-f" "../genero.4gs/chk_cncl.src"
"FORMCOLORS=fourgen" "ROWID_TYPE=integer" "002.tmp"
In the first one, you'll see that the last arg is missing on the called
side.
The 2nd call is clearly showing junk being passed across, as well as the
same last argument missing. It looks very much like the sort of strings I'd
expect to find at the end of my string region, since they're the strings
I'm working with.
The 3rd one also has a junk arg ("ac0001.tmp") which happens to pass the
command line rules for the called program, so it just does it's thing.
The 4th and final call fails the whole process because the junk "002.tmp" on
the command line causes the called program to complain and everything
terminates. If it was all working properly, I'd expect another 20 odd calls
to the nested program as it processes all the files in the directory.
In the file fposix.ri at about line 1145, I see a very telling comment:
/*
* we are subverting the RTT type subsystem here w.r.t. garbage
* collection but we're going to be doing an exec() so ...
*/
Without having gone much further into the code than that, I'm going to guess
that, since it won't be doing exec() on Windows, this subversion is guilty
in some way?
Now that I'm in this far, I'd have to study the rest of the code before I
can comment further, but maybe I've already given enough information for
someone to slap their forehead and fix it?
For the record, I've just tested a tiny little test program on UNIX and it
doesn't appear to be losing my last argument, but when copied across to
Windows, the last argument "c" disappears. This dumb program is not showing
junk, but I barely tried to do anything clever enough to cause more string
allocations - my little scan context is useless due to lack of foresight.
caller.icn:
procedure main()
cmd := ["callee", "a", "b", "c"]
system(cmd)
"this is a string" ? {
while move(1) do move(1)
}
system(cmd)
end
callee.icn:
procedure main(argv)
writes("CALLEE")
every writes(" ", image(!argv))
write()
end
------------------------------------------------------------------------------
Crystal Reports - New Free Runtime and 30 Day Trial
Check out the new simplified licensing option that enables unlimited
royalty-free distribution of the report engine for externally facing
server and web deployment.
http://p.sf.net/sfu/businessobjects
_______________________________________________
Unicon-group mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/unicon-group
