Tiny Core Linux
Tiny Core Base => TCB Bugs => Topic started by: webb on August 11, 2019, 02:56:35 AM
-
webb@box ~ filetool.sh -b
Backing up files to /mnt/vda1/tce/mydata.tgz/
Done.caused by echo -n on line 173 of filetool.sh combined with a following line outputting a slash, maybe sudo /bin/tar -C /
-
Hi webb
The extra / character is coming from the rotdash (rotating dash) command. The rotdash command writes the following sequence:
/ Backspace
- Backspace
\ Backspace
| BackspaceSo depending on how long your backup takes, any one of those characters could be left behind. The attached patch file fixes it
by including a leading space character in the next 2 echo commands. That will overwrite the character left by rotdash.
[EDIT]: Moved attachment to the next post. Rich
-
Hi Juanito
The attached patch file does the following:
1. Adds a trailing space to the "Backing up files to $MOUNTPOINT/$FULLPATH/${MYDATA}.tgz" message so that rotdash does not
print right next to the trailing z. This also provides a space if the "encrypting .. " message gets printed which itself will overwrite
anything that rotdash leaves behind.
2. Adds a leading space to the "\nDone." message to overwrite anything that rotdash leaves behind.
-
Hi Juanito
If there's any interest, I have a patch for rotdash that leaves the screen as it found it when it exits. Compiling under TC4 and
using sstrip increases the executables size from 1688 bytes to 1708 bytes. In TC10 I can't get the size under 12324 bytes.
The addition line of code is shown here with a comment and the diff is attached below:
#include <stdio.h>
#include <unistd.h>
int main(int argc, char **argv){
if (argc != 2) return 1;
const char arr[] = { '/', '-', '\\', '|' };
unsigned int i = 0;
chdir("/proc");
while (access(argv[1],F_OK) == 0) {
printf("%c\b",arr[i]);
fflush(stdout);
i++;
i %= 4;
usleep(100000);
}
// This clears from the cursor to the end of the line leaving
// the screen and cursor as they were when the program started.
printf("\033[K");
return 0;
}
The source code I used came from here:
http://distro.ibiblio.org/tinycorelinux/4.x/x86/release/src/3.x/rotdash.c
-
I often output rotdash into a <textarea> on a web page via filetool.sh. The backspace does not work now but it isn't a big problem.
If you add an escape character into the output it would be really bad for displaying via a web page.
I already have replacement rotdash code ready if it is decided you decide to procede.
Example:
[ INFO ] Backing up files to /mnt/mmcblk0p2/tce/mydata.tgz/-\|/ Done.
[ OK ] Backup successful.
-
Let's go with the filetool patch, given Greg's web use.
-
tinycore git updated
-
Hi Greg Erskine
I often output rotdash into a <textarea> on a web page via filetool.sh. The backspace does not work now but it isn't a big problem.
If you add an escape character into the output it would be really bad for displaying via a web page. ...
I had no idea anyone used this for a web page. I suppose I could have printed a space followed by a backspace:
printf(" \b");Since the backspace moves the cursor back 1 position the trailing space would simply get overwritten if more text were added.
If the next character was a newline, it would print the trailing space which wouldn't be visible anyway. This was the mechanism
that made the trailing slash visible.
-
Compiling under TC4 and using sstrip increases the executables size from 1688 bytes to 1708 bytes. In TC10 I can't get the size under 12324 bytes.
Hi, Rich!
Sorry for offtopic. Can you kindly explain to the newbie such a size difference?
Not a first mentioning of TC4 at the forum as the kind of milestone. What have changed later? Some kernel bloat?
-
Hi jazzbiker
... Can you kindly explain to the newbie such a size difference? ...
I honestly don't know why the executable is so much larger. My guess is the current compiler has some "enhancements" not present
in the TC4 version that add some overhead. There appear to be some differences in how they were built.
TC4:
tc@box:~/rotdash$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/tmp/tcloop/gcc/usr/local/bin/../lib/gcc/i686-pc-linux-gnu/4.6.1/lto-wrapper
Target: i686-pc-linux-gnu
Configured with: ../gcc-4.6.1/configure --prefix=/usr/local --libexecdir=/usr/local/lib --enable-shared --enable-threads=posix --enable-__cxa_atexit --enable-clocale=gnu --enable-languages=c,c++ --disable-multilib --disable-bootstrap --with-system-zlib --enable-frame-pointer
Thread model: posix
gcc version 4.6.1 (GCC)
tc@box:~/rotdash$
TC10:
tc@E310:~/rotdash$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/tmp/tcloop/gcc/usr/local/bin/../lib/gcc/i486-pc-linux-gnu/8.2.0/lto-wrapper
Target: i486-pc-linux-gnu
Configured with: ../configure --prefix=/usr/local --enable-languages=c,c++ --disable-multilib --disable-bootstrap --disable-libmpx --with-system-zlib --libexecdir=/usr/local/lib --enable-frame-pointer --with-mpfr=/usr/local --with-gmp=/usr/local
Thread model: posix
gcc version 8.2.0 (GCC)
tc@E310:~/rotdash$
-
My guess is the current compiler has some "enhancements" not present
in the TC4 version that add some overhead.
Thanks for Your answer, Rich.
My feeling is that sometimes tcc is the real alternative to gcc - for simple programs.
-
Continuing off-topic:
tc@box:/tmp$ uname -a
Linux box 4.19.10-tinycore #1999 SMP Tue Dec 18 13:36:47 UTC 2018 i686 GNU/Linux
tc@box:/tmp$ wget http://distro.ibiblio.org/tinycorelinux/4.x/x86/release/src/3.
x/rotdash.c
Connecting to distro.ibiblio.org (152.19.134.43:80)
rotdash.c 100% |********************************| 317 0:00:00 ETA
tc@box:/tmp$ tce-load -i tcc
tcc is already installed!
tc@box:/tmp$ tcc -v
tcc version 0.9.26 (i386 Linux)
tc@box:/tmp$ tcc -o rotdash rotdash.c
tc@box:/tmp$ ls -l rotdash*
-rwxr-xr-x 1 tc staff 2816 Aug 13 13:49 rotdash
-rw-r--r-- 1 tc staff 317 Aug 13 13:48 rotdash.c
tc@box:/tmp$ tce-load -i sstrip
sstrip is already installed!
tc@box:/tmp$ sstrip rotdash
tc@box:/tmp$ ls -l rotdash*
-rwxr-xr-x 1 tc staff 1312 Aug 13 13:49 rotdash
-rw-r--r-- 1 tc staff 317 Aug 13 13:48 rotdash.c
-
Hi jazzbiker
I honestly don't know why the executable is so much larger. My guess is the current compiler has some "enhancements" not present
in the TC4 version that add some overhead. ...
I don't know why it's happening, but I know what's happening. I looked through the stripped file and found 3 large blocks of zeros.
Each block was over 3500 bytes long.
... In TC10 I can't get the size under 12324 bytes. ...
Subtracting 3 times 3500 from 12324 leaves 1824 which suggests all the growth consists of zero padding.
-
I don't know why it's happening, but I know what's happening. I looked through the stripped file and found 3 large blocks of zeros.
Each block was over 3500 bytes long.
Some initialized variables. They may be examined better before stripping, probably.
But tcc gives 1312 total size )
-
tc@box:/tmp/rotdash$ nm -n -S rotdash
U access@@GLIBC_2.0
U chdir@@GLIBC_2.0
U fflush@@GLIBC_2.0
w __gmon_start__
U __libc_start_main@@GLIBC_2.0
U printf@@GLIBC_2.0
U usleep@@GLIBC_2.0
08049000 T _init
08049090 T _start
080490c0 00000004 T __x86.get_pc_thunk.bx
080490c4 t deregister_tm_clones
080490ec t register_tm_clones
08049122 t __do_global_dtors_aux
08049140 t frame_dummy
08049142 000000bb T main
080491fd 0000004f T __libc_csu_init
0804924c 00000001 T __libc_csu_fini
08049250 T _fini
0804a000 00000004 R _fp_hw
0804a004 00000004 R _IO_stdin_used
0804a014 r __FRAME_END__
0804bf04 t __frame_dummy_init_array_entry
0804bf04 t __init_array_start
0804bf08 t __do_global_dtors_aux_fini_array_entry
0804bf08 t __init_array_end
0804bf0c d _DYNAMIC
0804c000 d _GLOBAL_OFFSET_TABLE_
0804c024 D __data_start
0804c024 W data_start
0804c028 D __dso_handle
0804c02c B __bss_start
0804c02c D _edata
0804c02c 00000004 B stdout@@GLIBC_2.0
0804c02c D __TMC_END__
0804c030 00000001 b completed.6835
0804c034 B _end
After _fini in text section next symbol is _fp_hw in read-only section and it is aligned on xxx000.
Then _GLOBAL_OFFSET_TABLE_ is aligned the same way.
There may be some linker switches concerning section alignment, i guess.
Th hole between __FRAME_END__ (read-only section) and __frame_dummy_init_array_entry (text section) is almost 8K.
-
Hi jazzbiker
It seems everytime Class transitions between Read Only Data (R or r) and Code (T or t) a large block of padding gets
added. The column to the right of Section is the number of bytes between the current line and the previous line. You can
see that number spikes at each transition:
"tc@E310:~/rotdash$ nm -n -S -t d -a -f sysv rotdash"
"Symbols from rotdash:"
Name Value Class Type Size Section
__gmon_start__ w NOTYPE *UND*
__libc_start_main@@GLIBC_2.0 U FUNC *UND*
access@@GLIBC_2.0 U FUNC *UND*
chdir@@GLIBC_2.0 U FUNC *UND*
fflush@@GLIBC_2.0 U FUNC *UND*
printf@@GLIBC_2.0 U FUNC *UND*
usleep@@GLIBC_2.0 U FUNC *UND*
0 a FILE *ABS*
.comment 0 n
crtstuff.c 0 a FILE *ABS*
crtstuff.c 0 a FILE *ABS*
init.c 0 a FILE *ABS*
rotdash.c 0 a FILE *ABS*
static-reloc.c 0 a FILE *ABS*
.interp 134513012 r
.note.ABI-tag 134513032 r 20
.hash 134513064 r 32
.gnu.hash 134513124 r 60
.dynsym 134513160 r 36
.dynstr 134513320 r 160
.gnu.version 134513430 r 110
.gnu.version_r 134513452 r 22
.rel.dyn 134513484 r 32
.rel.plt 134513500 r 16
.init 134516736 t 3236
_init 134516736 T FUNC .init 0
.plt 134516768 t 32
.text 134516880 t 112
main 134516880 T FUNC 140 .text 0
_start 134517020 T FUNC .text 140
__x86.get_pc_thunk.bx 134517056 T FUNC 4 .text 36
deregister_tm_clones 134517060 t FUNC .text 4
register_tm_clones 134517100 t FUNC .text 40
__do_global_dtors_aux 134517154 t FUNC .text 54
frame_dummy 134517184 t FUNC .text 30
__libc_csu_init 134517186 T FUNC 79 .text 2
__libc_csu_fini 134517265 T FUNC 1 .text 79
.fini 134517268 t 3
_fini 134517268 T FUNC .fini 0
.rodata 134520832 r 3564
_fp_hw 134520832 R OBJECT 4 .rodata 0
_IO_stdin_used 134520836 R OBJECT 4 .rodata 4
.eh_frame 134520852 r 16
__FRAME_END__ 134520852 r OBJECT .eh_frame 0
.init_array 134528772 t 7920
__frame_dummy_init_array_entry 134528772 t OBJECT .init_array 0
__init_array_start 134528772 t NOTYPE .init_array 0
.fini_array 134528776 t 4
__do_global_dtors_aux_fini_array_entry 134528776 t OBJECT .fini_array 0
__init_array_end 134528776 t NOTYPE .init_array 0
.dynamic 134528780 d 4
_DYNAMIC 134528780 d OBJECT .dynamic 0
.got 134529020 d 240
.got.plt 134529024 d 4
_GLOBAL_OFFSET_TABLE_ 134529024 d OBJECT .got.plt 0
.data 134529060 d 36
__data_start 134529060 D NOTYPE .data 0
data_start 134529060 W NOTYPE .data 0
__dso_handle 134529064 D OBJECT .data 4
.bss 134529068 b 4
__TMC_END__ 134529068 D OBJECT .data 0
__bss_start 134529068 B NOTYPE .bss 0
_edata 134529068 D NOTYPE .data 0
stdout@@GLIBC_2.0 134529068 B OBJECT 4 .bss 0
completed.6835 134529072 b OBJECT 1 .bss 4
_end 134529076 B NOTYPE .bss 4
-
Hi, Rich!
This gaps grows along with the gcc version numbers. I tried the same test with gcc 5.4 and found this gaps in the same places, but being smaller ones.
Gcc is so powerfull and versatile, using it for tiny programs is like stirring the sugar in a tea with the drill )
Your case opened my eyes on tcc, thanks!
-
Stirring sugar again
tc@box:/tmp$ gcc -c rotdash.c
tc@box:/tmp$ tcc -o rotdash rotdash.o
tc@box:/tmp$ sstrip rotdash
tc@box:/tmp$ ls -l rotdash*
-rwxr-xr-x 1 tc staff 1280 Aug 16 08:54 rotdash
-rw------- 1 tc staff 317 Aug 16 08:53 rotdash.c
-rw-r--r-- 1 tc staff 1156 Aug 16 08:54 rotdash.o
-
Hi jazzbiker
I did a little digging and I think I figured out some of the reasons for the larger executables. The default linker script changed.
From:
elf_i386.xc # Combines and sorts relocatable sections.
To:
elf_i386.xce # Combines and sorts relocatable sections with a separate code segment.
The formula used to compute gap sizes between code and data changed.
This is the result if you specify the elf_i386.xbn script (Mix code & data on same page; don't align data):
tc@E310:~/rotdash$ gcc -march=i486 -mtune=i686 -Os -pipe -Wall rotdash.c -o rotdash -Wl,-T/usr/local/lib/ldscripts/elf_i386.xbn
tc@E310:~/rotdash$ ls -l rotdash
-rwxr-xr-x 1 tc staff 4704 Aug 18 01:48 rotdash
tc@E310:~/rotdash$ sstrip rotdash
tc@E310:~/rotdash$ ls -l rotdash
-rwxr-xr-x 1 tc staff 1656 Aug 18 01:48 rotdash
tc@E310:~/rotdash$
-
Interesting - there's something I was not aware of..
-
Hi Juanito and jazzbiker
I did a further writeup on executable sizes here if anyones interested:
http://forum.tinycorelinux.net/index.php/topic,23623.0.html
-
Thanks for that - I recompiled rotdash for CorePure between tc-10.x and tc-11.x - there was a small reduction in size, presumably because the original was from 4.x..