/lib/arm-linux-gnueabihf/libc.so.6: version GLIBC_2.34 not found

[Fred vS]

Did you already had a chance to try the option that was suggested by MarcoV and/or try a cross-build ? (both might be worth a shot).

Not yet on the native rpi and for cross-build I have to study how to do it this week-end... 

[TRon]

Not yet on the native rpi and for cross-build I have to study how to do it this week-end... 

Ok, no problem.

Just take your time to understand how to do a cross-build. It is valuable knowledge (even though in practice it is used only once in a while).

[Fred vS]

Not yet on the native rpi and for cross-build I have to study how to do it this week-end... 

Ok, no problem.

Just take your time to understand how to do a cross-build. It is valuable knowledge (even though in practice it is used only once in a while).

I use the precious fpcupdeluxe to cross-compile since decades for all my apps but I was never able to cross-compile fpc src himself.  :'(

[Thaddy]

I can do that since rpi 1. maybe discuss in private?
The most important thing is getting the right toolchain.

[Fred vS]

I can do that since rpi 1. maybe discuss in private?
The most important thing is getting the right toolchain.

Or maybe it could be interesting for others too? 
About the toolchain, the one from magic fpcupdeluxe works perfectly here (for all my apps, apart fpc-src of course).

Even better, to cross-compile from Linux to MacOS no need to be annoyed to install XCode, fpcupdeluxe has all needed... 
But this is for fpc-apps, not for fpc-src himself.

[Fred vS]

Did you already had a chance to try the option that was suggested by MarcoV and/or try a cross-build ? (both might be worth a shot).

OK, I did try on native Rpi to compile fpc 3.2.4 + libc symbol table with OPT="-Aas" and there is still the same error from assembler. 

[EDIT] Did try on rpi with 2024-10-22-raspios-bullseye-arm64 but same error.

[Fred vS]

Some news from the front.

It's not just fpc-src that can't compile when assigning a symbol table.

This code doesn't compile on native rpi or with cross-compile:

Code: Pascal  [Select][+][-]

1. program testaarch64;
2. procedure free(__ptr:pointer);cdecl;external 'c' name 'free@GLIBC_2.17';
3. begin
4. free(nil);
5. end.

Cross-compiled from Linux amd64 with:

Code: Bash  [Select][+][-]

1. $ ppcrossa64 testaarch64.pas

Gives that error:

Code: Pascal  [Select][+][-]

1. Free Pascal Compiler version 3.2.3-1397-g22da36f83e [2024/07/30] for aarch64
2. Copyright (c) 1993-2024 by Florian Klaempfl and others
3. Target OS: Linux for AArch64
4. Compiling testaarch64.pas
5. Assembling testaarch64
6. testaarch64.s: Assembler messages:
7. testaarch64.s:16: Error: unexpected characters following instruction at operand 1 -- `bl free@GLIBC_2.17'
8. testaarch64.pas(5,4) Error: Error while assembling exitcode 1
9. testaarch64.pas(5,4) Fatal: There were 2 errors compiling module, stopping
10. Fatal: Compilation aborted

 

Note that this code with direct compile on Linux amd64:

Code: Pascal  [Select][+][-]

1. program testamd64;
2. procedure free(__ptr:pointer);cdecl;external 'c' name 'free@GLIBC_2.2.5';
3. begin
4. free(nil);
5. end.

Compiled with this is ok:

Code: Pascal  [Select][+][-]

1. $ fpc testamd64.pas

Note also that I was able to cross-compile fpc-source form Linux amd64 to aarch64 but get the same error when some libc methods have symbol table.

[TRon]

Yes, confirmed on pi native with 3.2.4rc1 Fred.

Code: ASM  [Select][+][-]

1.         .file "testaarch64.pas"
2. // Begin asmlist al_procedures
3.  
4. .section .text.n_main
5.         .balign 8
6. .globl  PASCALMAIN
7.         .type   PASCALMAIN,@function
8. PASCALMAIN:
9. .globl  main
10.         .type   main,@function
11. main:
12.         stp     x29,x30,[sp, #-16]!
13.         mov     x29,sp
14.         bl      fpc_initializeunits
15.         movz    x0,#0
16.         bl      free@GLIBC_2.17
17.         bl      fpc_do_exit
18.         ldp     x29,x30,[sp], #16
19.         ret
20. .Le0:
21.         .size   main, .Le0 - main
22.  
23. .section .text
24. // End asmlist al_procedures
25. // Begin asmlist al_globals
26.  
27. .section .data.n_INITFINAL
28.         .balign 8
29. .globl  INITFINAL
30.         .type   INITFINAL,@object
31. INITFINAL:
32.         .quad   1,0
33.         .quad   INIT$_$SYSTEM
34.         .quad   0
35. .Le1:
36.         .size   INITFINAL, .Le1 - INITFINAL
37.  
38. .section .data.n_FPC_THREADVARTABLES
39.         .balign 8
40. .globl  FPC_THREADVARTABLES
41.         .type   FPC_THREADVARTABLES,@object
42. FPC_THREADVARTABLES:
43.         .long   1
44.         .quad   THREADVARLIST_$SYSTEM$indirect
45. .Le2:
46.         .size   FPC_THREADVARTABLES, .Le2 - FPC_THREADVARTABLES
47.  
48. .section .data.n_FPC_RESOURCESTRINGTABLES
49.         .balign 8
50. .globl  FPC_RESOURCESTRINGTABLES
51.         .type   FPC_RESOURCESTRINGTABLES,@object
52. FPC_RESOURCESTRINGTABLES:
53.         .quad   0
54. .Le3:
55.         .size   FPC_RESOURCESTRINGTABLES, .Le3 - FPC_RESOURCESTRINGTABLES
56.  
57. .section .data.n_FPC_WIDEINITTABLES
58.         .balign 8
59. .globl  FPC_WIDEINITTABLES
60.         .type   FPC_WIDEINITTABLES,@object
61. FPC_WIDEINITTABLES:
62.         .quad   0
63. .Le4:
64.         .size   FPC_WIDEINITTABLES, .Le4 - FPC_WIDEINITTABLES
65.  
66. .section .data.n_FPC_RESSTRINITTABLES
67.         .balign 8
68. .globl  FPC_RESSTRINITTABLES
69.         .type   FPC_RESSTRINITTABLES,@object
70. FPC_RESSTRINITTABLES:
71.         .quad   0
72. .Le5:
73.         .size   FPC_RESSTRINITTABLES, .Le5 - FPC_RESSTRINITTABLES
74.  
75. .section .fpc.n_version
76.         .balign 16
77.         .type   __fpc_ident,@object
78. __fpc_ident:
79.         .ascii  "FPC 3.2.4-rc1 [2024/09/23] for aarch64 - Linux"
80. .Le6:
81.         .size   __fpc_ident, .Le6 - __fpc_ident
82.  
83. .section .data.n___stklen
84.         .balign 8
85. .globl  __stklen
86.         .type   __stklen,@object
87. __stklen:
88.         .quad   8388608
89. .Le7:
90.         .size   __stklen, .Le7 - __stklen
91.  
92. .section .data.n___heapsize
93.         .balign 8
94. .globl  __heapsize
95.         .type   __heapsize,@object
96. __heapsize:
97.         .quad   0
98. .Le8:
99.         .size   __heapsize, .Le8 - __heapsize
100.  
101. .section .data.n___fpc_valgrind
102.         .balign 8
103. .globl  __fpc_valgrind
104.         .type   __fpc_valgrind,@object
105. __fpc_valgrind:
106.         .byte   0
107. .Le9:
108.         .size   __fpc_valgrind, .Le9 - __fpc_valgrind
109.  
110. .section .data.n_FPC_RESLOCATION
111.         .balign 8
112. .globl  FPC_RESLOCATION
113.         .type   FPC_RESLOCATION,@object
114. FPC_RESLOCATION:
115.         .quad   0
116. .Le10:
117.         .size   FPC_RESLOCATION, .Le10 - FPC_RESLOCATION
118. // End asmlist al_globals
119. .section .note.GNU-stack,"",%progbits
120.  

edit:

Note that this code with direct compile on Linux amd64:

FWIW:

Code: Bash  [Select][+][-]

1. $ fpc -B -a testx86_64.pas
2. Free Pascal Compiler version 3.2.3 [2024/09/23] for x86_64
3. Copyright (c) 1993-2024 by Florian Klaempfl and others
4. Note: Switching assembler to default source writing assembler
5. Target OS: Linux for x86-64
6. Compiling testx86_64.pas
7. Assembling testx86_64
8. testx86_64.s: Assembler messages:
9. testx86_64.s:21: Error: junk `@GLIBC_2.2.5' after expression
10. Error: Error while assembling exitcode 1
11. Fatal: There were 2 errors compiling module, stopping
12. Fatal: Compilation aborted
13. Error: ppcx64 returned an error exitcode
14.  

[Fred vS]

Yes, confirmed on pi native with 3.2.4rc1 Fred.

Code: ASM  [Select][+][-]

1.         .file "testaarch64.pas"
2. // Begin asmlist al_procedures
3.  
4. .section .text.n_main
5.         .balign 8
6. .globl  PASCALMAIN
7.         .type   PASCALMAIN,@function
8. PASCALMAIN:
9. .globl  main
10.         .type   main,@function
11. main:
12.         stp     x29,x30,[sp, #-16]!
13.         mov     x29,sp
14.         bl      fpc_initializeunits
15.         movz    x0,#0
16.         bl      free@GLIBC_2.17
17.         bl      fpc_do_exit
18.         ldp     x29,x30,[sp], #16
19.         ret
20. .Le0:
21.         .size   main, .Le0 - main
22.  
23. .section .text
24. // End asmlist al_procedures
25. // Begin asmlist al_globals
26.  
27. .section .data.n_INITFINAL
28.         .balign 8
29. .globl  INITFINAL
30.         .type   INITFINAL,@object
31. INITFINAL:
32.         .quad   1,0
33.         .quad   INIT$_$SYSTEM
34.         .quad   0
35. .Le1:
36.         .size   INITFINAL, .Le1 - INITFINAL
37.  
38. .section .data.n_FPC_THREADVARTABLES
39.         .balign 8
40. .globl  FPC_THREADVARTABLES
41.         .type   FPC_THREADVARTABLES,@object
42. FPC_THREADVARTABLES:
43.         .long   1
44.         .quad   THREADVARLIST_$SYSTEM$indirect
45. .Le2:
46.         .size   FPC_THREADVARTABLES, .Le2 - FPC_THREADVARTABLES
47.  
48. .section .data.n_FPC_RESOURCESTRINGTABLES
49.         .balign 8
50. .globl  FPC_RESOURCESTRINGTABLES
51.         .type   FPC_RESOURCESTRINGTABLES,@object
52. FPC_RESOURCESTRINGTABLES:
53.         .quad   0
54. .Le3:
55.         .size   FPC_RESOURCESTRINGTABLES, .Le3 - FPC_RESOURCESTRINGTABLES
56.  
57. .section .data.n_FPC_WIDEINITTABLES
58.         .balign 8
59. .globl  FPC_WIDEINITTABLES
60.         .type   FPC_WIDEINITTABLES,@object
61. FPC_WIDEINITTABLES:
62.         .quad   0
63. .Le4:
64.         .size   FPC_WIDEINITTABLES, .Le4 - FPC_WIDEINITTABLES
65.  
66. .section .data.n_FPC_RESSTRINITTABLES
67.         .balign 8
68. .globl  FPC_RESSTRINITTABLES
69.         .type   FPC_RESSTRINITTABLES,@object
70. FPC_RESSTRINITTABLES:
71.         .quad   0
72. .Le5:
73.         .size   FPC_RESSTRINITTABLES, .Le5 - FPC_RESSTRINITTABLES
74.  
75. .section .fpc.n_version
76.         .balign 16
77.         .type   __fpc_ident,@object
78. __fpc_ident:
79.         .ascii  "FPC 3.2.4-rc1 [2024/09/23] for aarch64 - Linux"
80. .Le6:
81.         .size   __fpc_ident, .Le6 - __fpc_ident
82.  
83. .section .data.n___stklen
84.         .balign 8
85. .globl  __stklen
86.         .type   __stklen,@object
87. __stklen:
88.         .quad   8388608
89. .Le7:
90.         .size   __stklen, .Le7 - __stklen
91.  
92. .section .data.n___heapsize
93.         .balign 8
94. .globl  __heapsize
95.         .type   __heapsize,@object
96. __heapsize:
97.         .quad   0
98. .Le8:
99.         .size   __heapsize, .Le8 - __heapsize
100.  
101. .section .data.n___fpc_valgrind
102.         .balign 8
103. .globl  __fpc_valgrind
104.         .type   __fpc_valgrind,@object
105. __fpc_valgrind:
106.         .byte   0
107. .Le9:
108.         .size   __fpc_valgrind, .Le9 - __fpc_valgrind
109.  
110. .section .data.n_FPC_RESLOCATION
111.         .balign 8
112. .globl  FPC_RESLOCATION
113.         .type   FPC_RESLOCATION,@object
114. FPC_RESLOCATION:
115.         .quad   0
116. .Le10:
117.         .size   FPC_RESLOCATION, .Le10 - FPC_RESLOCATION
118. // End asmlist al_globals
119. .section .note.GNU-stack,"",%progbits
120.  

Note that the first post talk about /lib/arm-linux-gnueabihf/libc.so.6 and this is the arm cpu 32 bit.
And it that case assigning symbol table is ok.

Thanks to jump into this TRon.

Fre;D

[Fred vS]

edit:

Note that this code with direct compile on Linux amd64:

FWIW:
...

Ooops I did not see that.

Sorry but I don't understand what this means. 

with this on x86_64 it compiles and runs ok.

Code: Pascal  [Select][+][-]

1. $ fpc -B testx86_64.pas

But with this I get error at compil, like you:

Code: Pascal  [Select][+][-]

1. $ fpc -B -a testx86_64.pas

 

[marcov]

-a is with the external assembler. So apparently GAS doesn't grok your symbol stuff, while the FPC internal assembler either ignores or processes it.

[Fred vS]

-a is with the external assembler. So apparently GAS doesn't grok your symbol stuff, while the FPC internal assembler either ignores or processes it.

For Linux on i386, x86_64 and ARM32 the FPC internal assembler processes perfectly:

Compiled with "official" fpc:

Code: Pascal  [Select][+][-]

1. $ nm -D  testx86_64
2.                  U free@GLIBC_2.2.5
3.                  w __gmon_start__
4.                  U __libc_start_main@GLIBC_2.34

Compiled with "patched" fpc-ootb:

Code: Pascal  [Select][+][-]

1. $ nm -D  testx86_64
2.                  U free@GLIBC_2.2.5
3.                  w __gmon_start__
4.                  U __libc_start_main@GLIBC_2.2.5

[TRon]

-a is with the external assembler. So apparently GAS doesn't grok your symbol stuff, while the FPC internal assembler either ignores or processes it.

Since -Aelf is illegal option for aarch64 shouldn't the generated assembler use .symver declaration for the asm output ?

I mean, do you perhaps happen to known what the end-goal/approach is here (in case there is any) ? PD mentioned being in preference of using the linker to solve this but that requires mapping (at least that is how gcc (to my understanding) seems to approach it).

Or is the idea perhaps to support the same features for all the other available internal writers ?