Re: [PATCH v4 2/2] gdb: CTF support

[Simon Marchi <simark at simark dot ca>]

On 2019-10-04 4:23 p.m., Weimin Pan wrote:
> This patch adds the CTF (Compact Ansi-C Type Format) support in gdb.
> Two submissions on which this gdb work depends were posted earlier
> in May:
> 
>  * On the binutils mailing list - adding libctf which creates, updates,
>    reads, and manipulates the CTF data.
>  * On the gcc mailing list - expanding gcc to directly emit the CFT data
>    with a new command line option -gt.
> 
> CTF is a reduced form of debugging information whose main purpose is to
> describe the type of C entities such as structures, unions, typedefs and
> function arguments at the global scope only. It does not contain debug
> information about source lines, location expressions, or local variables.
> For more information on CTF, see the documentation in the libdtrace-ctf
> source tree, available here:
> 
> <https://raw.githubusercontent.com/oracle/libdtrace-ctf/master/doc/ctf-format>.
> 
> This patch expands struct elfinfo by adding the .ctf section, which
> contains CTF debugging info, and modifies elf_symfile_read() to read it.
> If both DWARF and CTF exist in a program, only DWARF will be read. CTF data
> will be read only when there is no DWARF. The two-stage symbolic reading
> and setting strategy, partial and full, was used.
> 
> File ctfread.c contains functions to transform CTF data into gdb's internal
> symbol table structures by iterately reading entries from CTF sections
> of "data objects", "function info", "variable info", and "data types"
> when setting up either partial or full symbol table. If the ELF symbol table
> is available, e.g. not stripped, the CTF reader will associate the found
> type information with these symbol entries. Due to the proximity between DWARF
> and CTF (CTF being a much simplified subset of DWARF), some DWARF implementation
> was reused to support CTF.
> 
> Test cases ctf-constvars.exp, ctf-cvexpr.exp, ctf-ptype.exp, and ctf-whatis.exp
> have been added to verify the correctness of this support.
> 
> This patch has missing features and limitations which we will add and
> address in the future patches.

Hi Weimin,

I got notified this morning that this patch series was pushed to master, although it
was not approved yet.  I don't think I had any remaining comments and was probably going
to approve this version anyway, but I am wondering if you pushed this version to master
by mistake instead of your personal users/ branch, since I told you to do so.

Please just fix up the ChangeLog entry below:

> diff --git a/gdb/ChangeLog b/gdb/ChangeLog
> index 3974823..03b9474 100644
> --- a/gdb/ChangeLog
> +++ b/gdb/ChangeLog
> @@ -1,5 +1,20 @@
>  2019-10-04  Weimin Pan  <weimin.pan@oracle.com>
>  
> +	* ctfread.c: New file.
> +	* ctfread.h: New file.
> +	* elfread.c: Include ctfread.h.
> +	(struct elfinfo text_p): New member ctfsect.
> +	(elf_locate_sections): Mark CTF section.
> +	(elf_symfile_read): Call elfctf_build_psymtabs.
> +	* Makefile.in (LIBCTF): Add.
> +	(CLIBS): Use it.
> +	(CDEPS): Likewise.
> +	(DIST): Add ctfread.c.
> +	* ../Makefile.def (dependencies): Add all-libctf to all-gdb
> +	* ../Makefile.in: Add "all-gdb: maybe-all-libctf"

As mentioned in the previous review, these last two lines must go in the top-level ChangeLog.

Simon