[Bug 326866] New: GCC: dwarf records for inlines appear incomplete - openSUSE Bugs - openSUSE Mailing Lists

newer
[Bug 231434] New: yast2 ca_mgm

[Bug 326866] New: GCC: dwarf records for inlines appear incomplete

older
[Bug 309367] New: USB PTP Class...

bugzilla_noreply＠novell.com

20 Sep 2007 20 Sep '07

15:59

https://bugzilla.novell.com/show_bug.cgi?id=326866#c301921 Summary: GCC: dwarf records for inlines appear incomplete Product: openSUSE 10.3 Version: RC 1 Platform: x86 OS/Version: Other Status: NEW Severity: Normal Priority: P5 - None Component: Development AssignedTo: rguenther@novell.com ReportedBy: jblunck@novell.com QAContact: qa@suse.de CC: matz@novell.com Found By: L3 This is a testcase for fate #301921. Description: SystemTap and other debugging tools use dwarf debuginfo to gather type information and storage locations of variables. This should work for all kernel code. The GCC isn't emitting complete debuginfo when functions are inlined. Since the kernel is heavily using inline functions its often not possible to generate tap scripts for this functions. References: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23551 http://sources.redhat.com/bugzilla/show_bug.cgi?id=1155 -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

Show replies by date

bugzilla_noreply＠novell.com

20 Sep 20 Sep

16:00

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c1 --- Comment #1 from Jan Blunck 2007-09-20 10:00:05 MST --- Created an attachment (id=173715) --> (https://bugzilla.novell.com/attachment.cgi?id=173715) Testcase namei.i -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

16:09

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c2 --- Comment #2 from Michael Matz 2007-09-20 10:09:28 MST --- To make life easier for us: please spell out exactly which inlined function you mean (inlined into where, i.e. which call), what debug information you expected, and what you instead got. I know we talked about this, but for reference please spell it out here. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

16:15

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c3 --- Comment #3 from Jan Blunck 2007-09-20 10:15:56 MST --- Oh, sorry. It is of course real_lookup(). Here you can see that e.g. name (<8194>) has no location list. <1><8176>: Abbrev Number: 66 (DW_TAG_subprogram) <8177> DW_AT_name : (indirect string, offset: 0x3b6f): real_lookup <817b> DW_AT_decl_file : 1 <817c> DW_AT_decl_line : 469 <817e> DW_AT_prototyped : 1 <817f> DW_AT_type : <1f38> <8183> DW_AT_inline : 1 (inlined) <8184> DW_AT_sibling : <81e6> <2><8188>: Abbrev Number: 64 (DW_TAG_formal_parameter) <8189> DW_AT_name : (indirect string, offset: 0x1694): parent <818d> DW_AT_decl_file : 1 <818e> DW_AT_decl_line : 468 <8190> DW_AT_type : <1f38> <2><8194>: Abbrev Number: 64 (DW_TAG_formal_parameter) <8195> DW_AT_name : (indirect string, offset: 0x338): name <8199> DW_AT_decl_file : 1 <819a> DW_AT_decl_line : 468 <819c> DW_AT_type : <2692> <2><81a0>: Abbrev Number: 54 (DW_TAG_formal_parameter) <81a1> DW_AT_name : nd <81a4> DW_AT_decl_file : 1 <81a5> DW_AT_decl_line : 468 <81a7> DW_AT_type : <25f4> <2><81ab>: Abbrev Number: 55 (DW_TAG_variable) <81ac> DW_AT_name : (indirect string, offset: 0x3b2): result <81b0> DW_AT_decl_file : 1 <81b1> DW_AT_decl_line : 470 <81b3> DW_AT_type : <1f38> <2><81b7>: Abbrev Number: 68 (DW_TAG_variable) <81b8> DW_AT_name : dir <81bc> DW_AT_decl_file : 1 <81bd> DW_AT_decl_line : 471 <81bf> DW_AT_type : <2318> <2><81c3>: Abbrev Number: 70 (DW_TAG_lexical_block) <81c4> DW_AT_sibling : <81d6> <3><81c8>: Abbrev Number: 55 (DW_TAG_variable) <81c9> DW_AT_name : (indirect string, offset: 0x263c): dentry <81cd> DW_AT_decl_file : 1 <81ce> DW_AT_decl_line : 490 <81d0> DW_AT_type : <1f38> <3><81d4>: Abbrev Number: 63 (DW_TAG_lexical_block) <2><81d6>: Abbrev Number: 70 (DW_TAG_lexical_block) <81d7> DW_AT_sibling : <81e4> <3><81db>: Abbrev Number: 56 (DW_TAG_lexical_block) <4><81dc>: Abbrev Number: 57 (DW_TAG_variable) <81dd> DW_AT_abstract_origin: <6efd> <4><81e1>: Abbrev Number: 63 (DW_TAG_lexical_block) <2><81e4>: Abbrev Number: 63 (DW_TAG_lexical_block) .. <2><82c1>: Abbrev Number: 93 (DW_TAG_inlined_subroutine) <82c2> DW_AT_abstract_origin: <8176> <82c6> DW_AT_ranges : 0x1f8 <82ca> DW_AT_call_file : 1 <82cb> DW_AT_call_line : 784 <3><82cd>: Abbrev Number: 76 (DW_TAG_formal_parameter) <82ce> DW_AT_abstract_origin: <81a0> <3><82d2>: Abbrev Number: 76 (DW_TAG_formal_parameter) <82d3> DW_AT_abstract_origin: <8194> <3><82d7>: Abbrev Number: 92 (DW_TAG_formal_parameter) <82d8> DW_AT_abstract_origin: <8188> <82dc> DW_AT_location : 0xd30 (location list) <3><82e0>: Abbrev Number: 85 (DW_TAG_lexical_block) <82e1> DW_AT_ranges : 0x218 <4><82e5>: Abbrev Number: 57 (DW_TAG_variable) <82e6> DW_AT_abstract_origin: <81ab> <4><82ea>: Abbrev Number: 86 (DW_TAG_variable) <82eb> DW_AT_abstract_origin: <81b7> <82ef> DW_AT_location : 0xd59 (location list) .. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

24 Sep 24 Sep

13:49

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c4 --- Comment #4 from Richard Guenther 2007-09-24 07:49:41 MST --- Created an attachment (id=174333) --> (https://bugzilla.novell.com/attachment.cgi?id=174333) non-working patch Patch against some version of gcc, introduces -fpreserve-function-arguments but misses provisions against propagation (may_propagate_copy needs adjustment I guess). This one could be made work for 4.1. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

28 Sep 28 Sep

14:45

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c5 Richard Guenther changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #174333|0 |1 is obsolete| | --- Comment #5 from Richard Guenther 2007-09-28 08:45:10 MST --- Created an attachment (id=175497) --> (https://bugzilla.novell.com/attachment.cgi?id=175497) patch fixing the problem up until var-tracking var-tracking can figure out all param locations with -Os -g -fpreserve-function-arguments on the testcase, but nothing ends up in the debuginfo still. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

14:49

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c6 --- Comment #6 from Richard Guenther 2007-09-28 08:49:43 MST --- Text size with -fpreserve-function-arguments is 15053 bytes vs. 14771 bytes without. Short testcase where we generate location lists for the parameters, but gdb does not show anything: static inline int foo(int i) { return i-1; } int foobar(int j) { return foo(j); } int main(int argc, char **argv) { return foobar(argc); } <1><162>: Abbrev Number: 5 (DW_TAG_subprogram) <163> DW_AT_sibling : <1b5> <167> DW_AT_external : 1 <168> DW_AT_name : (indirect string, offset: 0xc0): foobar <16c> DW_AT_decl_file : 1 <16d> DW_AT_decl_line : 7 <16e> DW_AT_prototyped : 1 <16f> DW_AT_type : <15b> <173> DW_AT_low_pc : 0x400498 <17b> DW_AT_high_pc : 0x40049c <183> DW_AT_frame_base : 2 byte block: 77 8 (DW_OP_breg7: 8) <2><186>: Abbrev Number: 6 (DW_TAG_formal_parameter) <187> DW_AT_name : j <189> DW_AT_decl_file : 1 <18a> DW_AT_decl_line : 6 <18b> DW_AT_type : <15b> <18f> DW_AT_location : 0 (location list) <2><193>: Abbrev Number: 7 (DW_TAG_inlined_subroutine) <194> DW_AT_abstract_origin: <140> <198> DW_AT_low_pc : 0x400498 <1a0> DW_AT_high_pc : 0x40049b <1a8> DW_AT_call_file : 1 <1a9> DW_AT_call_line : 8 <3><1aa>: Abbrev Number: 8 (DW_TAG_formal_parameter) <1ab> DW_AT_abstract_origin: <151> <1af> DW_AT_location : 0x36 (location list) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

15:14

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c8 --- Comment #8 from Richard Guenther 2007-09-28 09:14:50 MST --- The last problem is because in add_location_or_const_value_attribute () if we have a (even bad) location list, we use that and do not exec /* We couldn't get any rtl, so try directly generating the location description from the tree. */ descr = loc_descriptor_from_tree (decl); if (descr) { add_AT_location_description (die, attr, descr); return; } which would have produced the DW_OP_reg5. Still the location list seems to fancy for gdb... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

16:07

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c9 --- Comment #9 from Michael Matz 2007-09-28 10:07:33 MST --- Hmm, Jans testcase seems to work just fine now, with the patch, and this on top (otherwise it might access REG_EXPR on a MEM): --- combine.c.mm 2007-09-28 17:31:05.000000000 +0200 +++ combine.c 2007-09-28 17:38:33.000000000 +0200 @@ -1269,8 +1269,10 @@ can_combine_p (rtx insn, rtx i3, rtx pre || (flag_preserve_function_arguments && GET_CODE (PATTERN (i3)) == PARALLEL && GET_CODE (XVECEXP (PATTERN (i3), 0, 0)) == ASM_OPERANDS - && REG_USERVAR_P (dest) - && REG_EXPR (dest) != REG_EXPR (src))) + && (REG_P (dest) != REG_P (src) + || (REG_P (dest) + && REG_USERVAR_P (dest) + && REG_EXPR (dest) != REG_EXPR (src))))) return 0; /* DEST must either be a REG or CC0. */ With this and these command line I get correct output: % ../gcc/cc1 -fpreserve-function-arguments -Wall -Wundef -Wstrict-prototypes -Wno-trigraphs -fno-strict-aliasing -fno-common -O2 -msoft-float -mregparm=3 -freg-struct-return -mpreferred-stack-boundary=4 -march=i686 -ffreestanding -maccumulate-outgoing-args -fomit-frame-pointer -g -fno-stack-protector -Wdeclaration-after-statement -Wno-pointer-sign -quiet namei.i readelf -wi namei.o ... <1><820b>: Abbrev Number: 63 (DW_TAG_subprogram) <820c> DW_AT_sibling : <827b> <8210> DW_AT_name : (indirect string, offset: 0x3a0b): real_lookup .. <2><8362>: Abbrev Number: 86 (DW_TAG_inlined_subroutine) <8363> DW_AT_sibling : <83ee> <8367> DW_AT_abstract_origin: <820b> <836b> DW_AT_ranges : 0x190 <836f> DW_AT_call_file : 1 <8370> DW_AT_call_line : 784 <3><8372>: Abbrev Number: 73 (DW_TAG_formal_parameter) <8373> DW_AT_abstract_origin: <8235> <8377> DW_AT_location : 0xe45 (location list) <3><837b>: Abbrev Number: 73 (DW_TAG_formal_parameter) <837c> DW_AT_abstract_origin: <8229> <8380> DW_AT_location : 0xe58 (location list) <3><8384>: Abbrev Number: 73 (DW_TAG_formal_parameter) <8385> DW_AT_abstract_origin: <821d> <8389> DW_AT_location : 0xe76 (location list) So we have location lists for all three formal arguments for the inlined call. But it indeed still breaks when I remove -mregparm=3 from the compile arguments. And for Richard it doesn't seem to work at all, so something is definitely still fishy. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

19:02

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c10 --- Comment #10 from Michael Matz 2007-09-28 13:01:59 MST --- RTL propagates copies into the asms, hence we have exactly the same problem as on GIMPLE. It works a bit better because var-tracking is clever, and hence can work-around some of those propagation, but in the namei.i testcase the inner 'name' decl (the one for the formal parameter of the inlined function) was completely gone and had no location list associated. Even in the -mregparm=3 case it did have a location list, but an incomplete one (in particular it didn't overlap the special asm). So we generally also need to avoid propagation into these special asms on RTL, which the attached patch does. Now it seems to work. At least the location lists overlap the asms. Now the problem is, that in instructions where the inner parameter is known to be in a register the outer parameter isn't anymore, although in reality both are the same. Perhaps some deficiency in var-tracking. After renaming the 'name' parameter of real_lookup to '_name' the varloc notes look like so: (note 452 55 453 10 ( _name (expr_list:REG_DEP_TRUE (reg:SI 5 di [orig:60 _name] [60]) (const_int 0 [0x0]))) NOTE_INSN_VAR_LOCATION) (note 453 452 454 10 ( name (nil)) NOTE_INSN_VAR_LOCATION) (note 454 453 455 10 ( nd (nil)) NOTE_INSN_VAR_LOCATION) (note 455 454 56 10 ( nd (expr_list:REG_DEP_TRUE (reg:SI 4 si [orig:61 nd ] [61]) (const_int 0 [0x0]))) NOTE_INSN_VAR_LOCATION) Note how _name gets the place of name, which in turn is evicted from it. Same for 'nd' (these are not the same decls, just named the same). Theoretically this isn't necessary, as at least in this case they hold the same value, and hence can both be placed in that register. That might influence debuggability, or might not. Conceptually during that instruction the outer 'nd' and 'name' really isn't accessible. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

19:04

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c11 --- Comment #11 from Michael Matz 2007-09-28 13:04:01 MST --- Created an attachment (id=175569) --> (https://bugzilla.novell.com/attachment.cgi?id=175569) patch for the RTL side -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

1 Oct 1 Oct

11:42

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c12 Richard Guenther changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #175497|0 |1 is obsolete| | Attachment #175569|0 |1 is obsolete| | --- Comment #12 from Richard Guenther 2007-10-01 05:42:35 MST --- Created an attachment (id=175707) --> (https://bugzilla.novell.com/attachment.cgi?id=175707) final patch against gcc 4.1 Combined patch. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

12:00

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c13 Richard Guenther changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |jblunck@novell.com --- Comment #13 from Richard Guenther 2007-10-01 06:00:26 MST --- Testing required. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

11 Oct 11 Oct

12:44

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c14 Richard Guenther changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #175707|0 |1 is obsolete| | --- Comment #14 from Richard Guenther 2007-10-11 06:44:15 MST --- Created an attachment (id=177680) --> (https://bugzilla.novell.com/attachment.cgi?id=177680) really final patch -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

22 Nov 22 Nov

09:42

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c15 Jan Blunck changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|jblunck@novell.com | --- Comment #15 from Jan Blunck 2007-11-22 02:42:39 MST --- Tested the compiler and it works fine. Impact on the code size (numbers for the kernel HEAD only): text data bss dec hex filename 3256136 1743970 839448 5839554 591ac2 boot/vmlinux-2.6.24-rc2-git4-2-default 3215143 1733410 837288 5785841 5848f1 boot/vmlinux-2.6.24-rc2-git4-2-default This is a 1% increase in the overall code size of the kernel. I haven't measured any performance benchmarks yet but I don't expect massive changes. Besides that it would be good if the compiler still would detect empty function bodies with -fpreserve-function-arguments enables. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

09:44

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c16 --- Comment #16 from Jan Blunck 2007-11-22 02:44:08 MST --- Created an attachment (id=184339) --> (https://bugzilla.novell.com/attachment.cgi?id=184339) Another testcase with empty body -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

bugzilla_noreply＠novell.com

23 Nov 23 Nov

15:40

New subject: [Bug 326866] GCC: dwarf records for inlines appear incomplete

https://bugzilla.novell.com/show_bug.cgi?id=326866#c17 Richard Guenther changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution| |FIXED --- Comment #17 from Richard Guenther 2007-11-23 08:40:27 MST --- Fixed. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Reply

Sign in to reply online Use email software

6005

Age (days ago)

6069

Last active (days ago)

Download

16 comments

1 participants

tags

participants (1)

bugzilla_noreply＠novell.com