[Buildroot] [Bug 13576] New: Issues compiling buildroot for 5x86

[Buildroot] [Bug 13576] New: Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

            Bug ID: 13576
           Summary: Issues compiling buildroot for 5x86
           Product: buildroot
           Version: unspecified
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: blocker
          Priority: P5
         Component: Other
          Assignee: unassigned at buildroot.uclibc.org
          Reporter: ingineru_de_sistem at yahoo.com
                CC: buildroot at uclibc.org
  Target Milestone: ---

Tried to compile a buildroot environment to revive an old p1mmx machine.

1) when choosing MariaDB to provide MySQL, perl DBD::mysql and rsyslog have
trouble finding mysql.h, have to add -I/path/to/output/target/.../mysql.h to
them individually

2) syslinux hits umpteen double definitions when linking.
Finally need to override the linker to always add --alow-multiple-definition
always, otherwise it won't compile. (Not sure it works, though).
(Uboot is not an option on a PC, lilo is absent from buildroot even though it
would perfectly fit the need. Grub is not recommended when booting from SD
cards, it tends to erase stuff when attempting to write and not waiting for the
whole page to be rewritten)

3) for some reason, buildroot needs to download a rust binary, but refuses to
skip the hash if the hash is not in the hash file. Removing rust*hash fixes
this.

...and still going

If it matters, the host is a VM running Ubuntu 20.04 LTS x64 .
Buildroot is the last version available on git (git clone ....)


I guess not many people are into reviving P1MMX machines with the latest kernel
and libraries

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #1 from Peter Korsgaard <jacmet@uclibc.org> ---
Sorry to hear of your issues. You will need to provide more details for us to
help though, E.G. what is your Buildroot configuration and what are the exact
error messages?

We indeed don't have a lilo package, but feel free to submit one if you are
willing to maintain it. Given how old it is, I could imagine that newer
toolchain versions could cause breakage.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

ingineru_de_sistem at yahoo.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ingineru_de_sistem at yahoo.co
                   |                            |m

--- Comment #2 from ingineru_de_sistem at yahoo.com ---
Created attachment 8776
  --> https://bugs.busybox.net/attachment.cgi?id=8776&action=edit
.config used

This is the config file requested, compressed with xz (Bugzilla won't attach it
otherwise).

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #3 from ingineru_de_sistem at yahoo.com ---
This is the .config I used (see attached file).
Try to use it yourself on a Ubuntu 20.04 x64 VM as a host, to get a closer look
at the issues I found.


2.2) Also forgot about this one:  when compiling syslinux, it cannot find the
strlen function to link in file
..../buildroot/output/build/syslinux-6.03/libinstaller/syslxmod.c
Since it's the only thing used from <string.h>, it's easier to just comment out
/* #include <string.h> */
and enter a quick and dirty replacement for it:
#ifndef strlen
int strlen(char *s){
 int l=0;
 while (*s) { ++l;++s;}
 return l;
}
#endif

it seems to like it

4) someone else also noticed that buildroot doesn't build a gcc binary for the
targets, even though it does build other development tools.
Please note that it IS needed for anyone who needs to program on the target in
C, C++, perl, python, lua, fortran 2018+, etc. .
One can try musl, but it's not the same as GNU gcc.

5) I understand the connection with busybox, but I'd still like an option to
not use it on systems that can afford the drive space. If this option already
exists, please document it.

6) syslinux finally worked, but with very little documentation.
Here's how I did it:
a) install the mbr binary in the first 440 bytes of the hdd image.
b) partition the image, make a linux partition of the exact same size as
rootfs.ext2 then a small FAT16 partition. Make the FAT16 partition bootable.
c) find the 32 bit syslinux binary in the buildroot folder, install the 32 bit
libraries on the host Ubuntu 20.04, run the 32-bit syslinux to install in the
FAT16 partition
d) manually add bzImage to the FAT16 partition, then create a syslinux.cfg file
there (documentation on the Syslinux web site)

*Guessing* that if one has an older system that needs an upgrade and IF the
upgrade goes on a different disk, one could use the old LILO with the new
kernel and the new disk or partition, without syslinux. Haven't tried this yet.
The LILO documentation is on TLDP (watch out for the case when installing on a
hdd that is not in the same position as when it will boot; it is documented and
useful).

At any rate, syslinux is needed to boot on a naked VM (for testing the image)
without any older boot loader.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #4 from ingineru_de_sistem at yahoo.com ---
7) stopping the "transmission" service fails.
This is fully documented in here:
https://forum.transmissionbt.com/viewtopic.php?t=11561

I tried that solution and it works fine (file /etc/init.d/S92transmission) : 

# diff -u S92transmission~ S92transmission
--- S92transmission~
+++ S92transmission
@@ -93,7 +93,7 @@
         #   1 if daemon was already stopped
         #   2 if daemon could not be stopped
         #   other if a failure occurred
-        start-stop-daemon --stop --quiet --retry=TERM/10/KILL/5 --pidfile
$PIDFILE --name $NAME
+        start-stop-daemon --stop --quiet --retry=TERM/10/KILL/5 --pidfile
$PIDFILE --exec $DAEMON
         RETVAL="$?"
         [ "$RETVAL" = 2 ] && return 2


(I mean it starts, stops without error and if I check the process afterwards it
is indeed starter or stopped as requested.)

It may be an upstream bug, but a simple patch in buildroot won't hurt it either
(if they fix it upstream, the patch will just skip, no biggie).

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #5 from Peter Korsgaard <jacmet@uclibc.org> ---
You are reporting a bunch of more-or-less unrelated issues and suggested
improvements, making it kind of hard to keep track of. Care to send your
suggestions as patches to the mailing list instead?

http://lists.busybox.net/mailman/listinfo/buildroot

Also for the bugreports, please be specific, E.G. you mentioned issues with
mariadb and syslinux. What is your exact defconfig and what is the exact error
message you get?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #6 from Peter Seiderer <ps.report@gmx.net> ---
Regarding Description - Point 3) - rust binary/missing hash:

Fixed by

https://patchwork.ozlabs.org/project/buildroot/patch/20210306112307.28337-1-ps.report at gmx.net/

The dependency is by package ripgrep --> host-rustc --> host-rust-bin (take a
look at 'make graph-depends' buildroot feature)...

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #7 from Peter Seiderer <ps.report@gmx.net> ---
Regarding Description - Point 3) - syslinux double definitions

Fixed (hopefully) by:

https://patchwork.ozlabs.org/project/buildroot/patch/20210307174753.9400-1-ps.report at gmx.net/

Out of interest, any references for the grub statement?


Regarding Description - Point 1) - MariaDB

No problem building your provided defconfig here, did you do a full
rebuild after changing your config (see [1] for details)?

In case a full rebuild fails, any special packages installed on your build
system?

[1] http://nightly.buildroot.org/manual.html#full-rebuild

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #8 from Peter Seiderer <ps.report@gmx.net> ---
Regarding Comment 4 Point 6) - syslinux installation

Take a look at board/technologic/ts5500/readme.txt for a similar system
configuration/setup?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #9 from ingineru_de_sistem at yahoo.com ---
(In reply to Peter Seiderer from comment #7)

About the grub issue:
For an older project of mine (reviving an old AMD Sempron laptop where the
original hdd had unrecoverable bad sectors), I tried Lubuntu 18.04 LTS with an
IDE44pin-to-SDCard convertor with a fairly new micro-SD card.
All went fine except for the part when booting it with grub would erase the
partition table at every boot from the hdd (well, sd-card+adapter).
After tons of Googling (can't remember the exact search terms) found that one
guy had replaced grub with lilo (which fortunately still comes as a Lubuntu
package at least in 18.04) and it worked. So I did the same (full install, add
the lilo package, purge grub, remove all grubby remnants, configure and run
lilo) and it worked fine.


For syslinux - if one wraps the linker in a shell file that adds
--allow-multiple-definition, all works fine. And it's even simpler than LILO
but it needs a FAT partition (no biggie), tested that one as well.

For the revival project at hand (ancient 32-bit 5*86 with an asterisk zap card)
I reused the ancient LILO which can load new kernels as well. 
Still a lot to do (Asterisk+zaptel doesn't compile directly either from
buildroot on 586, GCC doesn't come at all with buildroot and GCC's
documentation is not very friendly as to what sources one needs to download to
do the whole bootstraping stuff) and since it's a pet low-priority project, it
will take a long while. If anything pops up in the mean time, I will let you
know.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #10 from ingineru_de_sistem at yahoo.com ---
(In reply to Peter Korsgaard from comment #5)

For the .config file, see previous attachment.

The issues with mariadb is that parl-dbd-mysql and rsyslog have trouble finding
mysql.h. Modifying Makefile and re-running make fixes that temporarily, so they
can finally compile.
Have NOT tested with Oracle's flavor.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #11 from ingineru_de_sistem at yahoo.com ---
(In reply to ingineru_de_sistem from comment #10)
oops:  s/parl/perl/

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Buildroot] [Bug 13576] Issues compiling buildroot for 5x86

[bugzilla at busybox.net]

https://bugs.busybox.net/show_bug.cgi?id=13576

--- Comment #12 from Peter Seiderer <ps.report@gmx.net> ---
(In reply to ingineru_de_sistem from comment #4)

Regarding sysv init script for transmission

Fixed by:

 
https://patchwork.ozlabs.org/project/buildroot/patch/20210310223743.7879-1-ps.report at gmx.net/

-- 
You are receiving this mail because:
You are on the CC list for the bug.