[PATCH] D55261: [llvm-dwarfdump] - Simplify the test case.
George Rimar via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Dec 4 09:09:55 PST 2018
grimar marked an inline comment as done.
grimar added inline comments.
================
Comment at: test/tools/llvm-dwarfdump/X86/no_debug_addr.s:5-6
+## When .debug_addr is absent for some reason, check we do not print a bogus section name.
+## We had a bug that could be a reason of that when
+## DW_RLE_startx_length/DW_RLE_base_addressx/DW_RLE_offset_pair tags were used.
----------------
dblaikie wrote:
> grimar wrote:
> > dblaikie wrote:
> > > Not sure I understand this comment - but I think I get where you're going with it, might just be some grammar improvement to make it more clear?
> > >
> > > Oh, actually, here's a thought - did this bug pop-up when using rnglists in .dwo files? They certainly won't have a debug_addr & would then be a fully realistic scenario rather than one where a section of debug info would have to be dropped to demonstrate the failure?
> > >
> > > So maybe a comment like this:
> > >
> > > ## Ensure bogus empty section names are not printed when dumping rnglists that reference debug_addr when it is not present (such as in .dwo files).
> > >
> > > Yep, looks like it does, so here's a really short example I worked up:
> > >
> > > .section .debug_info.dwo,"e", at progbits
> > > .long .Ldebug_info_dwo_end1-.Ldebug_info_dwo_start1 # Length of Unit
> > > .Ldebug_info_dwo_start1:
> > > .short 5 # DWARF version number
> > > .byte 5 # DWARF Unit Type
> > > .byte 8 # Address Size (in bytes)
> > > .long 0 # Offset Into Abbrev. Section
> > > .quad -6809755978868859807
> > > .byte 1 # Abbrev [1] 0x14:0x32 DW_TAG_compile_unit
> > > .byte 0 # DW_AT_ranges
> > > .Ldebug_info_dwo_end1:
> > > .section .debug_abbrev.dwo,"e", at progbits
> > > .byte 1 # Abbreviation Code
> > > .byte 17 # DW_TAG_compile_unit
> > > .byte 0 # DW_CHILDREN_no
> > > .byte 85 # DW_AT_ranges
> > > .byte 35 # DW_FORM_rnglistx
> > > .byte 0 # EOM(1)
> > > .byte 0 # EOM(2)
> > > .section .debug_rnglists.dwo,"e", at progbits
> > > .long .Ldebug_rnglist_table_end1-.Ldebug_rnglist_table_start1 # Length
> > > .Ldebug_rnglist_table_start1:
> > > .short 5 # Version
> > > .byte 8 # Address size
> > > .byte 0 # Segment selector size
> > > .long 1 # Offset entry count
> > > .Lrnglists_dwo_table_base0:
> > > .long .Ldebug_ranges0-.Lrnglists_dwo_table_base0
> > > .Ldebug_ranges0:
> > > .byte 1 # DW_RLE_base_addressx
> > > .byte 0 # base address index
> > > .byte 4 # DW_RLE_offset_pair
> > > .byte 0 # starting offset
> > > .byte 1 # ending offset
> > > .byte 3 # DW_RLE_startx_length
> > > .byte 1 # start index
> > > .byte 2 # length
> > > .byte 0 # DW_RLE_end_of_list
> > > .Ldebug_rnglist_table_end1:
> > >
> > >
> > >
> > Idea about using the .dwo is interesting, thanks! I'll check it and probably update this patch tomorrow.
> >
> > > Oh, actually, here's a thought - did this bug pop-up when using rnglists in .dwo files?
> >
> > I faced this in LLD actually. Its implementation of the `llvm::DWARFObject` did not override the `getRnglistsSection`
> > and the effect from returning the empty section by default implementation was the same as rnglists section is absent.
> Guess you meant "getAddrSection"/addr section is absent - yeah, could totally believe that - great find! & the same bug totally shows up in other places, like the dwo dumping - so handy to fix all around :)
Yeah, it was `getAddrSection`, I looked at the wrong place.
(Had to add both `getRnglistsSection` and `getAddrSection` in that patch: https://reviews.llvm.org/D55109)
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D55261/new/
https://reviews.llvm.org/D55261
More information about the llvm-commits
mailing list