Date: Sat, 3 Sep 1994 11:36:26 +0200
Alternate-Recipient: Allowed
From: HANS ELLENBERGER <ELE AT CLIENTS DOT SWITCH DOT ch>
To: djgpp <djgpp AT sun DOT soe DOT clarkson DOT edu>
Subject: maint1/4 bugs


Having downloaded maint1/4 I can confirm that libc now perfectly
compiles. 

My current buglist holds:

a) system("tst1.bat arg1") still crashes.

b) go32_dpmi_remaining_physical_memory() is wrong when malloced()
memory pages have not yet been accessed and after spawnxxx().

Since it does not reflect free'd memory in the pools of malloc.c,
I suggest to keep code as is and to update documentation
accordingly.

c) stubedit make.exe keepmem=4m applied to newly generated make
3.67 seems not to have any effect. According to topline, make
uses approx. 3M which all are swapped out when spawning gcc.exe.

Setting gcc.exe's keepmem to 4M does not change anything either.

Anyhow, the programs now work without crashing under XMS, which
is a big improvement.

d) while (spawnlp(P_WAIT,"mem.exe","mem.exe",0L) == 0) often crashes
when control c or control break is pressed. Somtimes the keyboard
handler seems to lock and only hardware reset stops the loop. 

e) When environment contains redirection like GO32=2s1,
compilation of the GNU compilers is not possible because the
promptline generated by go32 interferes with the programs newly
sonnetized by make.

I suggest to output that prompt only when the commandline is
empty or when it contains the parameter -v.

f) makefile for cc1plus has no optimizing CFLAGS set.