README and other cosmetic updates

This commit is contained in:
Gregory Nutt 2014-01-04 08:38:05 -06:00
parent dedfdb982a
commit 84596e8473
2 changed files with 103 additions and 1 deletions

View File

@ -120,6 +120,108 @@ nsh
Windows build, but I would expect the same issues as is listed
for the ostest configuration..
STATUS:
Currently, NSH failes nsh_consoleoutput(). Here is an example.
This echo command causes the system to hang:
nsh> echo abc
Below is some annotated output from the debugger. Here is the 30,000 ft view:
- cmd_echo loops for each argv[i], 1 >=i > argc.
- It calls:
vtbl->output(vtbl, "%s ", argv[i])
where the prototype for output is:
int (*output)(FAR struct nsh_vtbl_s *vtbl, const char *fmt, ...);
- vtbl->output maps to nsh_consoleoutput() in this case.
- cmd_echo passes all of the arguments to output in registers.
- nsh_consoleoutput expects all of the parameters on the stack.
- nsh_console calls vfprintf() using bad values from the stack.
- vfprintf crashes and never returns.
Looks like a compiler bug to me.
# int cmd_echo(FAR struct nsh_vtbl_s *vtbl, int argc, char **argv)
#
# All input parameters are in registers
#
# R1=00802DA0 # vtbl
# R2=00000002 # argc
# R3=00802D15 # argv
# # argv[0]=00802DD7
# # argv[1]=00802DDC
# # 00802DD7="echo\0abc\0"
# SP=00802CDD
0001024C 05F0 PUSHMHI <R8-R11> # SP=00802CCD
0001024E 4418 LD R8,R1 # R8=00802DA0 vtbl
00010250 442A LD R10,R2 # R10=00000002 argc
00010252 443B LD R11,R3 # R11=00802D15 argv
00010254 3901 LD R9,#%1 # R9=00000001 arg index
00010256 C00C JP %10270
00010270 A5A9 CP R9,R10 # Bottom of loop
00010272 E1F2 JP lt,%10258
00010258 48840010 LD R4,%10(R8) # R4=00011156 adddress of output() method
0001025C 4490 LD R0,R9 # R0=00000001 Index of argv[1]
0001025E BC20 SLL R0,#%2 # R0=00000004 Offset to argv[1]
00010260 A0B0 ADD R0,R11 # R0=00802D19 Address of argv[1]
00010262 4481 LD R1,R8 # R1=00802DA0 vtbl address
00010264 452200008ADB LD R2,#%8ADB # R2=00008ADB = "%s "
0001026A 1203 LD R3,(R0) # R3=00802DDC Value of argv[1]
0001026C F214 CALL (R4) # Call vtbl->output(vtbl, "%s ", argv[i]);
# vtbl->output is nsh_consoleoutput
# static int nsh_consoleoutput(FAR struct nsh_vtbl_s *vtbl, const char *fmt, ...)
#
# All parameters are in registers:
#
# R1=00802DA0 vtbl address
# R2=00008ADB "%s "
# R3=00802DDC Value of argv[1]
# First is a check if the output file is open
#
# if (nsh_openifnotopen(pstate) != 0)
# {
# return ERROR;
# }
00011156 0800 LINK #%0 # SP=00802CC9, R14=00802CC9
00011158 5C81 LD R1,%8(FP) # R1=0000017F Should be value file FILE * for output
0001115A DF96 CALL %11088 # Call nsh_openifnotopen(), returns R0=00000000
0001115C 9000 CP R0,#%0
0001115E E602 JP z,%11164 # Skip over error return
00011160 30FF LD R0,#-%1
00011162 C007 JP %11172
# Then the failing call to vfprintf:
#
# va_start(ap, fmt);
# ret = vfprintf(pstate->cn_outstream, fmt, ap);
# va_end(ap);
#
# return ret;
00011164 4D03 LEA R3,%10(FP) # R3=00802CD5 ap=GARBAGE
00011166 5C80 LD R0,%8(FP) # R0=0000017F Should be value of pstate
00011168 48010033 LD R1,%33(R0) # R1=01000000 pstate->cn_outstream. Looks suspicious
0001116C 5CC2 LD R2,%C(FP) # R2=00802DA0
0001116E F10003FB CALL %11968 # Call vfprintf(01000000, 00802DA0, 00802CD5)
# All arguments are bad
# Does not survive call to vfprintf
ostest
------

View File

@ -53,7 +53,7 @@
* Private Functions
****************************************************************************/
static inline char *getfilepath(const char *name)
static inline char *getfilepath(const char *name)
{
snprintf(line, PATH_MAX, "%s/" DEFCONFIG, name);
line[PATH_MAX] = '\0';