From vineash30640 at gmail.com  Thu May 11 15:48:28 2023
From: vineash30640 at gmail.com (=?UTF-8?B?546L5Yqg55uK?=)
Date: Thu, 11 May 2023 22:48:28 +0800
Subject: [klee-dev] get the absolute path of file in InstructionInfo
Message-ID: <CAFPjMJWjetF1X-LDbWxcuyYtvVQCqD-uNdN89ky1bUuUDtOL8w@mail.gmail.com>

Hi,

I have a question about the `const std::string &file` in `struct
InstructionInfo`. From my observation, this `file` is always a relative
path, regardless of whether the file is in klee-uclibc, klee runtime, or my
own testing subjects. For example, we have:

   - runtime/POSIX/fd.c
   - ../lib/getopt.c
   - libc/string/strcmp.c

This makes it difficult to analyze, especially in `run.istats`, as tools
like kcachegrind cannot find the corresponding source code with relative
paths. Is there a way to make the `file` absolute?

Thanks,

Jiayi
-------------- next part --------------
HTML attachment scrubbed and removed

From f.busse at imperial.ac.uk  Sun May 21 11:55:01 2023
From: f.busse at imperial.ac.uk (Frank Busse)
Date: Sun, 21 May 2023 11:55:01 +0100
Subject: [klee-dev] get the absolute path of file in InstructionInfo
In-Reply-To: <CAFPjMJWjetF1X-LDbWxcuyYtvVQCqD-uNdN89ky1bUuUDtOL8w@mail.gmail.com>
References: <CAFPjMJWjetF1X-LDbWxcuyYtvVQCqD-uNdN89ky1bUuUDtOL8w@mail.gmail.com>
Message-ID: <20230521115501.19a0680d@haengemotte.localdomain>

Hi Jiayi,


On Thu, 11 May 2023 22:48:28 +0800
??? <vineash30640 at gmail.com> wrote:

> I have a question about the `const std::string &file` in `struct
> InstructionInfo`. From my observation, this `file` is always a
> relative path, regardless of whether the file is in klee-uclibc, klee
> runtime, or my own testing subjects. For example, we have:
> 
>    - runtime/POSIX/fd.c
>    - ../lib/getopt.c
>    - libc/string/strcmp.c
> 
> This makes it difficult to analyze, especially in `run.istats`, as
> tools like kcachegrind cannot find the corresponding source code with
> relative paths. Is there a way to make the `file` absolute?

Either add a field for the directory to that struct (the directory is
in the debug info, you can see it at the end of assembly.ll) or
configure your kcachegrind to access those directories. I don't have it
at hand currently but it's sth. like View -> Settings -> Directory
Settings for Source Annotation.


Kind regards,

Frank


From mcsinyx at disroot.org  Tue May 23 08:38:25 2023
From: mcsinyx at disroot.org (=?utf-8?q?Nguy=E1=BB=85n_Gia_Phong?=)
Date: Tue, 23 May 2023 16:38:25 +0900
Subject: [klee-dev] Use of -sym-stdin/stdout
In-Reply-To: <2b05ec25-3219-c2a7-3c27-1a0275c52a85@imperial.ac.uk>
Message-ID: <CSTHI51KGKYM.24WAHDG6EZ5B7@guix>

On Tue Feb 9 11:33:00 GMT 2021, Cristian Cadar wrote:
> printf is run concretely by default (you can add -DKLEE_SYM_PRINTF 
> when compiling KLEE's uclibc to change this)

Do you recall the reason for that default,
i.e. what are the potential drawbacks of enabling
symbolic printf in KLEE's ?Clibc?


From f.busse at imperial.ac.uk  Tue May 23 18:05:00 2023
From: f.busse at imperial.ac.uk (Frank Busse)
Date: Tue, 23 May 2023 18:05:00 +0100
Subject: [klee-dev] Use of -sym-stdin/stdout
In-Reply-To: <CSTHI51KGKYM.24WAHDG6EZ5B7@guix>
References: <2b05ec25-3219-c2a7-3c27-1a0275c52a85@imperial.ac.uk>
	<CSTHI51KGKYM.24WAHDG6EZ5B7@guix>
Message-ID: <20230523180500.1ce839d5@gyali>

Hi,


On Tue, 23 May 2023 16:38:25 +0900
Nguy?n Gia Phong <mcsinyx at disroot.org> wrote:

> Do you recall the reason for that default,
> i.e. what are the potential drawbacks of enabling
> symbolic printf in KLEE's ?Clibc?

printf has to transform its parameters into strings. This is heavily
branching code (even worse when the format string is symbolic). Have a
look e.g. at musl's implementation:

https://git.musl-libc.org/cgit/musl/tree/src/stdio/vfprintf.c


Kind regards,

Frank


From mcsinyx at disroot.org  Wed May 24 08:24:15 2023
From: mcsinyx at disroot.org (=?utf-8?q?Nguy=E1=BB=85n_Gia_Phong?=)
Date: Wed, 24 May 2023 16:24:15 +0900
Subject: [klee-dev] Use of -sym-stdin/stdout
In-Reply-To: <20230523180500.1ce839d5@gyali>
References: <2b05ec25-3219-c2a7-3c27-1a0275c52a85@imperial.ac.uk>
	<CSTHI51KGKYM.24WAHDG6EZ5B7@guix> <20230523180500.1ce839d5@gyali>
Message-ID: <CSUBTW6ONKLA.29GIP2X19JSRF@guix>

On 2023-05-23 at 18:05+01:00, Frank Busse wrote:
> On Tue, 23 May 2023 16:38:25 +0900
> Nguy?n Gia Phong <mcsinyx at disroot.org> wrote:
> > what are the potential drawbacks of enabling
> > symbolic printf in KLEE's ?Clibc?
>
> printf has to transform its parameters into strings. This is heavily
> branching code (even worse when the format string is symbolic). Have a
> look e.g. at musl's implementation:
>
> https://git.musl-libc.org/cgit/musl/tree/src/stdio/vfprintf.c

Thanks, that explains it for me.