Building FPC 3.2.2 on Debian "Testing" (Bookworm) x86_64

[TRon]

Original issue:bootstrap=3.0.2 fpc=3.0.4 (default binutils 2.40) wrt msg2inc segmentation fault.

Either stripped or not "file" tells it is a valid executable on both accounts. Invoking it (outside the build process) displays the segmentation fault.

Code: [Select]

make DEBUG=1 OPT="-O- -k--verbose=5" PP="${pp_bootstrap_compiler}"

-- snip --

make -C utils cleanall
make[6]: Entering directory '/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/compiler/utils'
/usr/bin/rm -f fpc ppufiles ppudump ppumove mkarmins mkx86ins fpc.o ppufiles.o ppudump.o ppumove.o mkarmins.o mkx86ins.o libpfpc.a libpppufiles.a libpppudump.a libpppumove.a libpmkarmins.a libpmkx86ins.a libimpfpc.a libimpppufiles.a libimpppudump.a libimpppumove.a libimpmkarmins.a libimpmkx86ins.a
/usr/bin/rm -f units/x86_64-linux/ppu.ppu units/x86_64-linux/crc.ppu
/usr/bin/rm -rf units
/usr/bin/rm -f *.o *.ppu *.rst *.s *.a *.so *.ppl
/usr/bin/rm -rf *.sl
/usr/bin/rm -f fpcmade.* Package.fpc ./ppas.sh script.res link.res 
/usr/bin/rm -f *_ppas.sh
make[6]: Leaving directory '/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/compiler/utils'
/usr/bin/rm -rf x86_64/units
/usr/bin/rm -f x86_64/*.o x86_64/*.ppu x86_64/*.rst x86_64/*.s x86_64/*.a x86_64/*.so x86_64/*.ppl
/usr/bin/rm -f x86_64/ppc386 x86_64/ppc68k x86_64/ppcx64 x86_64/ppcppc x86_64/ppcsparc x86_64/ppcppc64 x86_64/ppcarm x86_64/ppcmips x86_64/ppcmipsel x86_64/ppcjvm x86_64/ppc8086 x86_64/ppcx64
/usr/bin/rm -f ppcx64
/usr/bin/mkdir -p x86_64/units/x86_64-linux
make ./msg2inc
make[6]: Entering directory '/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/compiler'
/media/ramdisk-8gb/compilehostedfpc/bin/ppcx64-3.0.2 -gl -Ur -Xs -O2 -n -Fux86_64 -Fusystems -Fu/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux -Fix86_64 -FE. -FUx86_64/units/x86_64-linux -Cg -dDEBUG -dRELEASE -O- -k--verbose=5 -dx86_64 -dGDB -dBROWSERLOG -Fux86 -Sew -FE. utils/msg2inc.pp
GNU ld (GNU Binutils for Debian) 2.40
  Supported emulations:
   elf_x86_64
   elf32_x86_64
   elf_i386
   elf_iamcu
   i386pep
   i386pe
using internal linker script:
==================================================
/* Script for -z combreloc -z separate-code */
/* Copyright (C) 2014-2023 Free Software Foundation, Inc.
   Copying and distribution of this script, with or without modification,
   are permitted in any medium without royalty provided the copyright
   notice and this notice are preserved.  */
OUTPUT_FORMAT("elf64-x86-64", "elf64-x86-64",
      "elf64-x86-64")
OUTPUT_ARCH(i386:x86-64)
ENTRY(_start)
SEARCH_DIR("=/usr/local/lib/x86_64-linux-gnu"); SEARCH_DIR("=/lib/x86_64-linux-gnu"); SEARCH_DIR("=/usr/lib/x86_64-linux-gnu"); SEARCH_DIR("=/usr/lib/x86_64-linux-gnu64"); SEARCH_DIR("=/usr/local/lib64"); SEARCH_DIR("=/lib64"); SEARCH_DIR("=/usr/lib64"); SEARCH_DIR("=/usr/local/lib"); SEARCH_DIR("=/lib"); SEARCH_DIR("=/usr/lib"); SEARCH_DIR("=/usr/x86_64-linux-gnu/lib64"); SEARCH_DIR("=/usr/x86_64-linux-gnu/lib");
SECTIONS
{
  PROVIDE (__executable_start = SEGMENT_START("text-segment", 0x400000)); . = SEGMENT_START("text-segment", 0x400000) + SIZEOF_HEADERS;
  .interp         : { *(.interp) }
  .note.gnu.build-id  : { *(.note.gnu.build-id) }
  .hash           : { *(.hash) }
  .gnu.hash       : { *(.gnu.hash) }
  .dynsym         : { *(.dynsym) }
  .dynstr         : { *(.dynstr) }
  .gnu.version    : { *(.gnu.version) }
  .gnu.version_d  : { *(.gnu.version_d) }
  .gnu.version_r  : { *(.gnu.version_r) }
  .rela.dyn       :
    {
      *(.rela.init)
      *(.rela.text .rela.text.* .rela.gnu.linkonce.t.*)
      *(.rela.fini)
      *(.rela.rodata .rela.rodata.* .rela.gnu.linkonce.r.*)
      *(.rela.data .rela.data.* .rela.gnu.linkonce.d.*)
      *(.rela.tdata .rela.tdata.* .rela.gnu.linkonce.td.*)
      *(.rela.tbss .rela.tbss.* .rela.gnu.linkonce.tb.*)
      *(.rela.ctors)
      *(.rela.dtors)
      *(.rela.got)
      *(.rela.bss .rela.bss.* .rela.gnu.linkonce.b.*)
      *(.rela.ldata .rela.ldata.* .rela.gnu.linkonce.l.*)
      *(.rela.lbss .rela.lbss.* .rela.gnu.linkonce.lb.*)
      *(.rela.lrodata .rela.lrodata.* .rela.gnu.linkonce.lr.*)
      *(.rela.ifunc)
    }
  .rela.plt       :
    {
      *(.rela.plt)
      PROVIDE_HIDDEN (__rela_iplt_start = .);
      *(.rela.iplt)
      PROVIDE_HIDDEN (__rela_iplt_end = .);
    }
  .relr.dyn : { *(.relr.dyn) }
  . = ALIGN(CONSTANT (MAXPAGESIZE));
  .init           :
  {
    KEEP (*(SORT_NONE(.init)))
  }
  .plt            : { *(.plt) *(.iplt) }
.plt.got        : { *(.plt.got) }
.plt.sec        : { *(.plt.sec) }
  .text           :
  {
    *(.text.unlikely .text.*_unlikely .text.unlikely.*)
    *(.text.exit .text.exit.*)
    *(.text.startup .text.startup.*)
    *(.text.hot .text.hot.*)
    *(SORT(.text.sorted.*))
    *(.text .stub .text.* .gnu.linkonce.t.*)
    /* .gnu.warning sections are handled specially by elf.em.  */
    *(.gnu.warning)
  }
  .fini           :
  {
    KEEP (*(SORT_NONE(.fini)))
  }
  PROVIDE (__etext = .);
  PROVIDE (_etext = .);
  PROVIDE (etext = .);
  . = ALIGN(CONSTANT (MAXPAGESIZE));
  /* Adjust the address for the rodata segment.  We want to adjust up to
     the same address within the page on the next page up.  */
  . = SEGMENT_START("rodata-segment", ALIGN(CONSTANT (MAXPAGESIZE)) + (. & (CONSTANT (MAXPAGESIZE) - 1)));
  .rodata         : { *(.rodata .rodata.* .gnu.linkonce.r.*) }
  .rodata1        : { *(.rodata1) }
  .eh_frame_hdr   : { *(.eh_frame_hdr) *(.eh_frame_entry .eh_frame_entry.*) }
  .eh_frame       : ONLY_IF_RO { KEEP (*(.eh_frame)) *(.eh_frame.*) }
  .sframe         : ONLY_IF_RO { *(.sframe) *(.sframe.*) }
  .gcc_except_table   : ONLY_IF_RO { *(.gcc_except_table .gcc_except_table.*) }
  .gnu_extab   : ONLY_IF_RO { *(.gnu_extab*) }
  /* These sections are generated by the Sun/Oracle C++ compiler.  */
  .exception_ranges   : ONLY_IF_RO { *(.exception_ranges*) }
  /* Adjust the address for the data segment.  We want to adjust up to
     the same address within the page on the next page up.  */
  . = DATA_SEGMENT_ALIGN (CONSTANT (MAXPAGESIZE), CONSTANT (COMMONPAGESIZE));
  /* Exception handling  */
  .eh_frame       : ONLY_IF_RW { KEEP (*(.eh_frame)) *(.eh_frame.*) }
  .sframe         : ONLY_IF_RW { *(.sframe) *(.sframe.*) }
  .gnu_extab      : ONLY_IF_RW { *(.gnu_extab) }
  .gcc_except_table   : ONLY_IF_RW { *(.gcc_except_table .gcc_except_table.*) }
  .exception_ranges   : ONLY_IF_RW { *(.exception_ranges*) }
  /* Thread Local Storage sections  */
  .tdata   :
   {
     PROVIDE_HIDDEN (__tdata_start = .);
     *(.tdata .tdata.* .gnu.linkonce.td.*)
   }
  .tbss   : { *(.tbss .tbss.* .gnu.linkonce.tb.*) *(.tcommon) }
  .preinit_array    :
  {
    PROVIDE_HIDDEN (__preinit_array_start = .);
    KEEP (*(.preinit_array))
    PROVIDE_HIDDEN (__preinit_array_end = .);
  }
  .init_array    :
  {
    PROVIDE_HIDDEN (__init_array_start = .);
    KEEP (*(SORT_BY_INIT_PRIORITY(.init_array.*) SORT_BY_INIT_PRIORITY(.ctors.*)))
    KEEP (*(.init_array EXCLUDE_FILE (*crtbegin.o *crtbegin?.o *crtend.o *crtend?.o ) .ctors))
    PROVIDE_HIDDEN (__init_array_end = .);
  }
  .fini_array    :
  {
    PROVIDE_HIDDEN (__fini_array_start = .);
    KEEP (*(SORT_BY_INIT_PRIORITY(.fini_array.*) SORT_BY_INIT_PRIORITY(.dtors.*)))
    KEEP (*(.fini_array EXCLUDE_FILE (*crtbegin.o *crtbegin?.o *crtend.o *crtend?.o ) .dtors))
    PROVIDE_HIDDEN (__fini_array_end = .);
  }
  .ctors          :
  {
    /* gcc uses crtbegin.o to find the start of
       the constructors, so we make sure it is
       first.  Because this is a wildcard, it
       doesn't matter if the user does not
       actually link against crtbegin.o; the
       linker won't look for a file to match a
       wildcard.  The wildcard also means that it
       doesn't matter which directory crtbegin.o
       is in.  */
    KEEP (*crtbegin.o(.ctors))
    KEEP (*crtbegin?.o(.ctors))
    /* We don't want to include the .ctor section from
       the crtend.o file until after the sorted ctors.
       The .ctor section from the crtend file contains the
       end of ctors marker and it must be last */
    KEEP (*(EXCLUDE_FILE (*crtend.o *crtend?.o ) .ctors))
    KEEP (*(SORT(.ctors.*)))
    KEEP (*(.ctors))
  }
  .dtors          :
  {
    KEEP (*crtbegin.o(.dtors))
    KEEP (*crtbegin?.o(.dtors))
    KEEP (*(EXCLUDE_FILE (*crtend.o *crtend?.o ) .dtors))
    KEEP (*(SORT(.dtors.*)))
    KEEP (*(.dtors))
  }
  .jcr            : { KEEP (*(.jcr)) }
  .data.rel.ro : { *(.data.rel.ro.local* .gnu.linkonce.d.rel.ro.local.*) *(.data.rel.ro .data.rel.ro.* .gnu.linkonce.d.rel.ro.*) }
  .dynamic        : { *(.dynamic) }
  .got            : { *(.got) *(.igot) }
  . = DATA_SEGMENT_RELRO_END (SIZEOF (.got.plt) >= 24 ? 24 : 0, .);
  .got.plt        : { *(.got.plt) *(.igot.plt) }
  .data           :
  {
    *(.data .data.* .gnu.linkonce.d.*)
    SORT(CONSTRUCTORS)
  }
  .data1          : { *(.data1) }
  _edata = .; PROVIDE (edata = .);
  . = .;
  __bss_start = .;
  .bss            :
  {
   *(.dynbss)
   *(.bss .bss.* .gnu.linkonce.b.*)
   *(COMMON)
   /* Align here to ensure that the .bss section occupies space up to
      _end.  Align after .bss to ensure correct alignment even if the
      .bss section disappears because there are no input sections.
      FIXME: Why do we need it? When there is no .bss section, we do not
      pad the .data section.  */
   . = ALIGN(. != 0 ? 64 / 8 : 1);
  }
  .lbss   :
  {
    *(.dynlbss)
    *(.lbss .lbss.* .gnu.linkonce.lb.*)
    *(LARGE_COMMON)
  }
  . = ALIGN(64 / 8);
  . = SEGMENT_START("ldata-segment", .);
  .lrodata   ALIGN(CONSTANT (MAXPAGESIZE)) + (. & (CONSTANT (MAXPAGESIZE) - 1)) :
  {
    *(.lrodata .lrodata.* .gnu.linkonce.lr.*)
  }
  .ldata   ALIGN(CONSTANT (MAXPAGESIZE)) + (. & (CONSTANT (MAXPAGESIZE) - 1)) :
  {
    *(.ldata .ldata.* .gnu.linkonce.l.*)
    . = ALIGN(. != 0 ? 64 / 8 : 1);
  }
  . = ALIGN(64 / 8);
  _end = .; PROVIDE (end = .);
  . = DATA_SEGMENT_END (.);
  /* Stabs debugging sections.  */
  .stab          0 : { *(.stab) }
  .stabstr       0 : { *(.stabstr) }
  .stab.excl     0 : { *(.stab.excl) }
  .stab.exclstr  0 : { *(.stab.exclstr) }
  .stab.index    0 : { *(.stab.index) }
  .stab.indexstr 0 : { *(.stab.indexstr) }
  .comment       0 : { *(.comment) }
  .gnu.build.attributes : { *(.gnu.build.attributes .gnu.build.attributes.*) }
  /* DWARF debug sections.
     Symbols in the DWARF debugging sections are relative to the beginning
     of the section so we begin them at 0.  */
  /* DWARF 1.  */
  .debug          0 : { *(.debug) }
  .line           0 : { *(.line) }
  /* GNU DWARF 1 extensions.  */
  .debug_srcinfo  0 : { *(.debug_srcinfo) }
  .debug_sfnames  0 : { *(.debug_sfnames) }
  /* DWARF 1.1 and DWARF 2.  */
  .debug_aranges  0 : { *(.debug_aranges) }
  .debug_pubnames 0 : { *(.debug_pubnames) }
  /* DWARF 2.  */
  .debug_info     0 : { *(.debug_info .gnu.linkonce.wi.*) }
  .debug_abbrev   0 : { *(.debug_abbrev) }
  .debug_line     0 : { *(.debug_line .debug_line.* .debug_line_end) }
  .debug_frame    0 : { *(.debug_frame) }
  .debug_str      0 : { *(.debug_str) }
  .debug_loc      0 : { *(.debug_loc) }
  .debug_macinfo  0 : { *(.debug_macinfo) }
  /* SGI/MIPS DWARF 2 extensions.  */
  .debug_weaknames 0 : { *(.debug_weaknames) }
  .debug_funcnames 0 : { *(.debug_funcnames) }
  .debug_typenames 0 : { *(.debug_typenames) }
  .debug_varnames  0 : { *(.debug_varnames) }
  /* DWARF 3.  */
  .debug_pubtypes 0 : { *(.debug_pubtypes) }
  .debug_ranges   0 : { *(.debug_ranges) }
  /* DWARF 5.  */
  .debug_addr     0 : { *(.debug_addr) }
  .debug_line_str 0 : { *(.debug_line_str) }
  .debug_loclists 0 : { *(.debug_loclists) }
  .debug_macro    0 : { *(.debug_macro) }
  .debug_names    0 : { *(.debug_names) }
  .debug_rnglists 0 : { *(.debug_rnglists) }
  .debug_str_offsets 0 : { *(.debug_str_offsets) }
  .debug_sup      0 : { *(.debug_sup) }
  .gnu.attributes 0 : { KEEP (*(.gnu.attributes)) }
  /DISCARD/ : { *(.note.GNU-stack) *(.gnu_debuglink) *(.gnu.lto_*) }
}


==================================================
/usr/bin/ld: mode elf_x86_64
attempt to open ./link.res succeeded
./link.res
opened script file ./link.res
/usr/bin/ld: warning: ./link.res contains output sections; did you forget -T?
attempt to open /media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/prt0.o succeeded
/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/prt0.o
attempt to open ./x86_64/units/x86_64-linux/msg2inc.o succeeded
./x86_64/units/x86_64-linux/msg2inc.o
attempt to open /media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/system.o succeeded
/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/system.o
attempt to open /media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/lnfodwrf.o succeeded
/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/lnfodwrf.o
attempt to open /media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/exeinfo.o succeeded
/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/exeinfo.o
attempt to open /media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/strings.o succeeded
/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/rtl/units/x86_64-linux/strings.o
make[6]: Leaving directory '/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/compiler'
./msg2inc msg/errore.msg msg msg
make[5]: *** [Makefile:3858: msgtxt.inc] Segmentation fault
make[5]: Leaving directory '/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/compiler'
make[4]: *** [Makefile:3971: next] Error 2
make[4]: Leaving directory '/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/compiler'
make[3]: *** [Makefile:3975: ppc1] Error 2
make[3]: Leaving directory '/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/compiler'
make[2]: *** [Makefile:3987: cycle] Error 2
make[2]: Leaving directory '/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4/compiler'
make[1]: *** [Makefile:2780: compiler_cycle] Error 2
make[1]: Leaving directory '/media/ramdisk-8gb/compilehostedfpc/src/fpc-3.0.4'
make: *** [Makefile:2812: build-stamp.x86_64-linux] Error 2

A 193mb log from LD_DEBUG=ALL is such overwhelming that I am unable to see a tree in the forest 

Next step is to check what building 3.2.2 shows us.


edit: it seems that msg2inc will not run without elevated rights (i am the owner, group owner and the executable has all required read/write rights). I have never seen that before and the source of msg2inc does not seem to justify that requirement (unless I missed something) . Anybody out there that is able to shed a light on how to further debug (gdb does not have a backtrace) ? Or is it simply the 'first call' to getrlimit that is failing when called without elevated rights ?

[MarkMLl]

Original issue:bootstrap=3.0.2 fpc=3.0.4 (default binutils 2.40) wrt msg2inc segmentation fault.

Either stripped or not "file" tells it is a valid executable on both accounts. Invoking it (outside the build process) displays the segmentation fault.

...

edit: it seems that msg2inc will not run without elevated rights (i am the owner, group owner and the executable has all required read/write rights). I have never seen that before and the source of msg2inc does not seem to justify that requirement (unless I missed something) . Anybody out there that is able to shed a light on how to further debug (gdb does not have a backtrace) ? Or is it simply the 'first call' to getrlimit that is failing when called without elevated rights ?

Basically, the "won't run without elevated rights" issue appears to apply to all versions of FPC older than 3.2 when running on Debian 12, and it also applies to any program built on Debian 12: not just intermediates when building the compiler.

The reason the compilers themselves in the examples below work is that they were built on Debian 11, hence this might- or might not- be related to the Symbol Versioning issues discussed in the long thread at https://forum.lazarus.freepascal.org/index.php/topic,64731.0.html

Code: Pascal  [Select][+][-]

1. program test;
2.  
3. begin
4. end.
5.  

Obviously if that runs it won't generate any output, but neither should it crash.

Code: Text  [Select][+][-]

1. ~$ fpc -V3.2.2 test.pas && ./test
2. Free Pascal Compiler version 3.2.2 [2022/10/24] for x86_64
3. Copyright (c) 1993-2021 by Florian Klaempfl and others
4. Target OS: Linux for x86-64
5. Compiling test.pas
6. Linking test
7. 5 lines compiled, 0.1 sec
8.  
9. ~$ fpc -V3.2.0 test.pas && ./test
10. Free Pascal Compiler version 3.2.0 [2022/10/24] for x86_64
11. Copyright (c) 1993-2020 by Florian Klaempfl and others
12. Target OS: Linux for x86-64
13. Compiling test.pas
14. Linking test
15. 5 lines compiled, 0.2 sec
16.  
17. ~$ fpc -V3.0.4 test.pas && ./test
18. Free Pascal Compiler version 3.0.4 [2022/10/24] for x86_64
19. Copyright (c) 1993-2017 by Florian Klaempfl and others
20. Target OS: Linux for x86-64
21. Compiling test.pas
22. Linking test
23. /usr/bin/ld: warning: link.res contains output sections; did you forget -T?
24. 5 lines compiled, 0.1 sec
25. Segmentation fault
26.  
27. ~$ sudo ./test
28. [sudo] password for markMLl:
29.  

This is obviously something which will not be fixed, but it would still be nice to work out what's going on since some older versions of FPC are worth having for testing purposes.

Is this behaviour documented anywhere?

MarkMLl

[MarkMLl]

Probably my last word on this unless anybody has any suggestions: it's "way above my pay grade".

Code: [Select]

Segmentation fault

Basically, this appears to be a completely different problem from the "symbol versioning" one, and is caused by the fact that the linker command has been reworked to- and I know there's much more to it than this- address the issue

Code: [Select]

/usr/bin/ld: warning: link.res contains output sections; did you forget -T?

This happened between FPC 3.0 and 3.2 as can be seen if one uses the -sh option to generate ppas.sh and link.res files. The later versions add the -T option (I think Jonas explained the reason it had been omitted to me at one point), and the result is a substantial change in output structure even though the linker version is the same.

If one wants a version of FPC for x86_64 in the range 2.6.4 through 3.0.4 to run on Debian 12, one has to build it on Debian 11. But even if it will run on Debian 12, the binaries it builds probably won't.

The error message might not even be the result of code generated by the compiler, which is why gdb can't get a grip on it. The structure of the executable binary is different between 3.0 and 3.2, and that might simply be the kernel's way of announcing that it's rejecting it for security reasons: which is why it will run if invoked using sudo etc.

MarkMLl