Bug 496423

Summary: something maybe busted in stdlib.h?
Product: [Fedora] Fedora Reporter: Tom Horsley <horsley1953>
Component: glibcAssignee: Jakub Jelinek <jakub>
Status: CLOSED NOTABUG QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: low    
Version: rawhideCC: jakub
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2009-04-23 10:06:42 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Attachments:
Description Flags
Buffer.cpp run through g++ -E
none
Buffer.cpp run through g++ -E -dD none

Description Tom Horsley 2009-04-19 00:43:48 UTC
Description of problem:

Trying to compile truecrypt on fedora 11, I get this:

Compiling Buffer.cpp
In file included from StringConverter.h:12,
                 from SerializerFactory.h:14,
                 from Serializable.h:16,
                 from Exception.h:14,
                 from Buffer.cpp:10:
/usr/include/stdlib.h:351: error: expected ‘;’ before ‘*’ token
/usr/include/stdlib.h:352: error: expected ‘;’ before ‘*’ token
/usr/include/stdlib.h:353: error: expected ‘;’ before ‘*’ token
/usr/include/stdlib.h:357: error: expected ‘;’ before ‘*’ token
/usr/include/stdlib.h:361: error: ‘int32_t’ has not been declared

I didn't have this problem on fedora 10, but I haven't yet analyzed
it enough to know if the source code is doing something weird or the
stdlib.h header is really busted.

Version-Release number of selected component (if applicable):
glibc-headers-2.9.90-16.x86_64
glibc-2.9.90-16.x86_64

How reproducible:
Every time I try to compile this code.

Steps to Reproduce:
1.way too complex to describe. This is version 5.0 of the truecrypt
source with some local patches applied.
2.
3.
  
Actual results:
error above

Expected results:
no compile problems

Additional info:

Comment 1 Tom Horsley 2009-04-19 01:13:51 UTC
Created attachment 340196 [details]
Buffer.cpp run through g++ -E

I'm not sure what the heck this code is doing. It appears to maybe be defining
int32 in terms of int32_t and vice-versa, but with some namespaces involved as well. I'm willing to call it a truecrypt source bug, but it is still true that
the same source compiles OK in fedora 10 (or did a while back, I suppose
recent updates may have busted it there as well).

I guess I need to hack the source even more to reliably compile.

Comment 2 Tom Horsley 2009-04-19 01:15:21 UTC
Here's the command line used by the Makefile to compile this. There
are lots of suspicous -D options as well (this is the version I modified
to use -E instead of -c):

g++ -E -MMD -I/zooty/build/truecrypt-5.0-source -I/zooty/build/truecrypt-5.0-source/Crypto -DBOOL=int -DFALSE=0 -DTRUE=1 -D__int8=char -D__int16=short -D__int32=int '-D__int64=long long'   -O2 -fno-strict-aliasing   -DTC_UNIX -DTC_LINUX -fdata-sections -ffunction-sections  -Wall -Wno-sign-compare -Wno-unused-parameter Buffer.cpp > Buffer.e

Comment 3 Jakub Jelinek 2009-04-22 09:36:36 UTC
Can you please add -dD to the options, so that macros can be seen in the preprocessed source as well?

Comment 4 Tom Horsley 2009-04-22 23:37:16 UTC
Created attachment 340850 [details]
Buffer.cpp run through g++ -E -dD

OK, here's the same file with -dD added to the command line.

I did discover a way to get this to work: After I unpack and patch the source,
I added a little script that runs through all the .cpp files and just
sticks:

#include <stdlib.h>

at the beginning of the file so it is included before all other local includes.

Everything seems to compile perfectly fine if I do that, and it is no worse
than most of the other hacks I'm doing to the source :-).

Comment 5 Jakub Jelinek 2009-04-23 10:06:42 UTC
That's clearly a truecrypt bug.
See how it does:
namespace TrueCrypt
{
#include <inttypes.h>
...
}
...
#include <stdlib.h>

That's invalid, standard headers may be only included outside of any external declaration or definition, so can't be included from within a namespace definition, see ISO C++ [lib.using.headers]/3.

Comment 6 Tom Horsley 2009-04-23 10:54:26 UTC
Yea, I was definitely leaning toward a truecrypt bug, I just couldn't
figure out what silly thing it was doing. Thanks for letting me know.