Re: [PATCH] Add replacement endian.h and byteswap.h to libgnu
* Ulf Hermann: > Some systems don't provide endian.h and byteswap.h. The required > functions are trivial to define using sys/param.h and gcc builtins, > though. > > Also, include endian.h in dwelf_scn_gnu_compressed_size.c as that uses > be64toh(). This is still an issue with non-glibc, non-BSD compilation. The patch is not ideal, as it depends on a GCC extension, but it's an improvement for those who use GCC on a platform which does not provide these functions as part of the C library.
Re: [PATCH] Add replacement endian.h and byteswap.h to libgnu
>> Some systems don't provide endian.h and byteswap.h. The required >> functions are trivial to define using sys/param.h and gcc builtins, >> though. >> >> Also, include endian.h in dwelf_scn_gnu_compressed_size.c as that uses >> be64toh(). > > This is still an issue with non-glibc, non-BSD compilation. The patch > is not ideal, as it depends on a GCC extension, but it's an > improvement for those who use GCC on a platform which does not provide > these functions as part of the C library. Well, there are a lot of other issues to be fixed if you want to build elfutils on anything non-gcc. We could add a further check for those builtins and sys/param.h and then add a somewhat less trivial version of endian.h and byteswap.h if they are missing. But that would likely involve other compiler extensions or OS-specific headers. br, Ulf
Re: [PATCH] Add replacement endian.h and byteswap.h to libgnu
* Ulf Hermann: >>> Some systems don't provide endian.h and byteswap.h. The required >>> functions are trivial to define using sys/param.h and gcc builtins, >>> though. >>> >>> Also, include endian.h in dwelf_scn_gnu_compressed_size.c as that uses >>> be64toh(). >> >> This is still an issue with non-glibc, non-BSD compilation. The patch >> is not ideal, as it depends on a GCC extension, but it's an >> improvement for those who use GCC on a platform which does not provide >> these functions as part of the C library. > > Well, there are a lot of other issues to be fixed if you want to build > elfutils on anything non-gcc. We could add a further check for those > builtins and sys/param.h and then add a somewhat less trivial version > of endian.h and byteswap.h if they are missing. But that would likely > involve other compiler extensions or OS-specific headers. Ah, fair enough. I mainly wanted to increase awareness of the patch because it has not been merged yet.
Re: [PATCH] tests: Add varlocs-self and exprlocs-self tests.
On Wed, 2017-11-15 at 21:01 +0100, Mark Wielaard wrote: > On Fri, 2017-11-10 at 17:12 +0100, Mark Wielaard wrote: > > Make sure the testcases (library functions they use) don't crash, > > triggers self-check/asserts or leaks memory under valgrind. This > > also helps making sure newer DWARF constructs are handled (when > > building with -gdwarf-5). > > I pushed this to master now. I setup a buildbot that checks all commits on various setups. It found an issue with this patch on ppc64/ppc64le: https://builder.wildebeest.org/buildbot/#/builders/11/builds/14 It was caused by a silly typo... I pushed the attached to master as obvious. Cheers, MarkFrom 1af6c2b9e49f59ea2f01035738525e798bb4f645 Mon Sep 17 00:00:00 2001 From: Mark Wielaard Date: Thu, 16 Nov 2017 13:20:11 +0100 Subject: [PATCH] tests: Fix cfi_debug => cfi_debug_bias typo in varlocs assert. We want to check whether the bias is zero, not whether we have dwarf debug_frame cfi. This triggered on a ppc64/ppc64le self-check since it has both debug_frame and eh_frame cfi (other arches often only have eh_frame). Signed-off-by: Mark Wielaard --- tests/ChangeLog | 4 tests/varlocs.c | 2 +- 2 files changed, 5 insertions(+), 1 deletion(-) diff --git a/tests/ChangeLog b/tests/ChangeLog index 6c8a8e1..0162034 100644 --- a/tests/ChangeLog +++ b/tests/ChangeLog @@ -1,3 +1,7 @@ +2017-11-16 Mark Wielaard + + * varlocs.c (main): Fix cfi_debug => cfi_debug_bias typo in assert. + 2017-11-10 Mark Wielaard * run-exprlocs-self.sh: New test. diff --git a/tests/varlocs.c b/tests/varlocs.c index 3267198..0e43229 100644 --- a/tests/varlocs.c +++ b/tests/varlocs.c @@ -1031,7 +1031,7 @@ main (int argc, char *argv[]) cfi_debug = dwfl_module_dwarf_cfi (mod, &cfi_debug_bias); cfi_eh = dwfl_module_eh_cfi (mod, &cfi_eh_bias); - assert (cfi_debug == 0); // No bias needed, same file. + assert (cfi_debug_bias == 0); // No bias needed, same file. // We are a bit forgiving for object files. There might be // relocations we don't handle that are needed in some -- 1.8.3.1
Re: [PATCH] Add replacement endian.h and byteswap.h to libgnu
On Thu, 2017-11-16 at 10:21 +0100, Florian Weimer wrote: > * Ulf Hermann: > > Well, there are a lot of other issues to be fixed if you want to build > > elfutils on anything non-gcc. We could add a further check for those > > builtins and sys/param.h and then add a somewhat less trivial version > > of endian.h and byteswap.h if they are missing. But that would likely > > involve other compiler extensions or OS-specific headers. > > Ah, fair enough. I mainly wanted to increase awareness of the patch > because it has not been merged yet. Yeah, sorry for dropping some of these older patches. I admit they aren't really on my radar anymore. So please do ping them if you want to get them integrated. But I believe we never fully finished the discussion about whether or not to adopt gnulib for things like this. Which I think is a better idea than having to maintain our own versions. Cheers, Mark
Re: [PATCH] readelf: Print DIE offset in attribute reading error messages.
On Tue, 2017-10-03 at 16:59 +0200, Mark Wielaard wrote: > When processing large files it is useful to know the DIE offset if > printing of attributes fails (especially when redirecting the output). > With this change the error message looks like: > > eu-readelf: DIE [2aeb8ef1] cannot get attribute value: invalid DWARF Got another report about a large debug file where this improved error message would have been really handy. So I pushed it to master now. Cheers, Mark
