1.40.4 incoming - ntg-pdftex - NTG Mailing Lists

newer
Xpdf 3.02 security hole (fwd)

1.40.4 incoming

older
Re: [NTG-pdftex] 1.40.4 incoming

Martin Schröder

7 Jul 2007 7 Jul '07

3:14 p.m.

Hi, I want to leave for TUG2007 with a released 1.40.4. The svn has in branches/stable the current version of 1.40.4. Please test and tell me what I've forgotten. :-) Best Martin

Reply

Sign in to reply online Use email software

Show replies by date

George N. White III

7 Jul 7 Jul

5:57 p.m.

On 7/7/07, Martin Schröder wrote:

I want to leave for TUG2007 with a released 1.40.4.

Thanks.

The svn has in branches/stable the current version of 1.40.4. Please test and tell me what I've forgotten. :-)

I just did: $ svn checkout svn://scm.foundry.supelec.fr/svn/pdftex/branches/stable [...] Checked out revision 171. [pdftex-1.40.4] I (slightly) abused your build.sh configuration by compiling with -std=c99 on Fedora Release 7 (i686, gcc (GCC) 4.1.2 20070502 (Red Hat 4.1.2-12)) and using system libraries (fewer chances for the compile to fail and less time to compile): --with-system-zlib \ --with-system-ncurses \ --with-system-pnglib \ --with-system-t1lib \ --with-system-gd \ --with-system-freetype \ This still gives the error for strsep: cc -march=i686 -std=c99 -DHAVE_CONFIG_H -I. -I../../../../src/texk/web2c/pdftexdir -I.. -I../../../../src/texk/web2c/pdftexdir/../.. -I.. -I../../../../src/texk/web2c/pdftexdir/.. -I../.. -I../../../../src/texk/web2c/pdftexdir/../.. -I../../../libs/md5 -I../../../../src/texk/web2c/pdftexdir/../../../libs/md5 -I../../../libs/xpdf -I../../../libs/xpdf/.. -I../../../../src/texk/web2c/pdftexdir/../../../libs/xpdf -I../../../../src/texk/web2c/pdftexdir/../../../libs/xpdf/.. -I../../../libs/obsdcompat -I../../../libs/obsdcompat/.. -I../../../../src/texk/web2c/pdftexdir/../../../libs/obsdcompat -I../../../../src/texk/web2c/pdftexdir/../../../libs/obsdcompat/.. -DpdfTeX -g -O2 -c ../../../../src/texk/web2c/pdftexdir/utils.c -o utils.o In file included from ../../../../src/texk/web2c/pdftexdir/utils.c:23: ../../../../src/texk/web2c/pdftexdir/../../../libs/obsdcompat/openbsd-compat.h:57: error: expected identifier or '(' before '__extension__' ../../../../src/texk/web2c/pdftexdir/../../../libs/obsdcompat/openbsd-compat.h:57: error: expected identifier or '(' before ')' token make[1]: *** [utils.o] Error 1 make[1]: Leaving directory `/Data/opt/tex/src/pdftex-1.40.4/source/build-i686-linux/texk/web2c/pdftexdir' make: *** [pdftexdir/libpdf.a] Error 2 My workaround is: mv libs/obsdcompat/config.h libs/obsdcompat/config.h.ac sed -e 's,/\* #undef HAVE_STRSEP \*/,#define HAVE_STRSEP 1 /* GNW */,' \ < libs/obsdcompat/config.h.ac \

libs/obsdcompat/config.h

I never got a response to my posting on the openssh list about this problem. The result passed the "fmtutil-sys --all" test and is able to compile story.tex and sample2.tex. I also checked that bug 762 is fixed: pdfTeX warning: pdftex: No flags specified for non-embedded font `Symbol' (psyr ) (I'm using 4): fix your map entry. [so the mapfiles shipped in TL2007 need to be adjusted] For those who haven't seen mention of linux-gate.so (it doesn't actually exist as a .so file http://www.trilithium.com/johan/2005/08/linux-gate/): [gwhite@cerberus pdftex-1.40.4]$ ldd $(which pdftex) linux-gate.so.1 => (0x00c89000) libpng12.so.0 => /usr/lib/libpng12.so.0 (0x004fd000) libz.so.1 => /lib/libz.so.1 (0x002cb000) libm.so.6 => /lib/libm.so.6 (0x00299000) libc.so.6 => /lib/libc.so.6 (0x00143000) /lib/ld-linux.so.2 (0x00124000) -- George N. White III Head of St. Margarets Bay, Nova Scotia

Reply

Sign in to reply online Use email software

Martin Schröder

6:42 p.m.

2007/7/7, George N. White III :

I (slightly) abused your build.sh configuration by compiling with -std=c99 on Fedora Release 7 (i686, gcc (GCC) 4.1.2 20070502 (Red Hat 4.1.2-12)) and using system libraries (fewer chances for the compile to fail and less time to compile):

Not that it should make a difference, but does the standard build.sh work? I have no problems with OpenSUSE 10.1 (gcc (GCC) 4.1.0 (SUSE Linux)) and 10.2 (gcc (GCC) 4.1.2 20061115 (prerelease) (SUSE Linux)) and -std=c99 -Wall. Best Martin

Reply

Sign in to reply online Use email software

George N. White III

9:02 p.m.

On 7/7/07, Martin Schröder wrote:

2007/7/7, George N. White III :

...
I (slightly) abused your build.sh configuration by compiling with -std=c99 on Fedora Release 7 (i686, gcc (GCC) 4.1.2 20070502 (Red Hat 4.1.2-12)) and using system libraries (fewer chances for the compile to fail and less time to compile):

Not that it should make a difference, but does the standard build.sh work?

Yes -- not tested, but the .pool files are identical between the standard build and my -std=c99 build.

I have no problems with OpenSUSE 10.1 (gcc (GCC) 4.1.0 (SUSE Linux)) and 10.2 (gcc (GCC) 4.1.2 20061115 (prerelease) (SUSE Linux)) and -std=c99 -Wall.

Interesting. Could it be that Red Hat has a different take on whether strsep is visible under std=c99? CC='cc -std=c99 -Wall' ./build.sh > build-c99-Wall.log 2>&1 failed at the same place, and from this build, source/build/libs/obsdcompat/config.log fails to detect strsep: configure:5077: checking whether strsep is declared configure:5106: cc -std=c99 -Wall -c -g -O2 -Wall -Wpointer-arith -Wuninitialized -Wsign-compare -Wno-pointer-sign conftest.c >&5 conftest.c: In function 'main': conftest.c:50: error: 'strsep' undeclared (first use in this function) conftest.c:50: error: (Each undeclared identifier is reported only once conftest.c:50: error: for each function it appears in.) conftest.c:50: warning: unused variable 'p' configure:5112: $? = 1 configure: failed program was: | /* confdefs.h. */ | | // ...shortened | #define HAVE_STRING_H 1 | //... | /* end confdefs.h. */ | | #ifdef HAVE_STRING_H | # include | #endif | | | int | main () | { | #ifndef strsep | char *p = (char *) strsep; | #endif | | ; | return 0; | } configure:5136: result: no ----------------------------------------------------------------------------- in the glibc changelog there is: * Wed Oct 03 2001 Jakub Jelinek 2.2.4-19 - fix strsep http://www.ecos.sourceware.org/ml/libc-hacker/2001-10/msg00001.html has some discussion of the patch. It may be that I need to add -D_BSD_SOURCE to get strsep on Red Hat, but I still don't like the test used by openssh's configure I'll try this with a different distribution (hopefully Monday). -- George N. White III Head of St. Margarets Bay, Nova Scotia

Reply

Sign in to reply online Use email software

George N. White III

10 Jul 10 Jul

7:07 p.m.

On 7/7/07, Martin Schröder wrote: small problem with SGI compilers, and I still don't understand finding strsep -- see below As for the prompt question -- a generic prompt as suggest by Philip Taylor is much better than something system-dependent. The same problem should lead to the same google search on all platforms so we don't end up with 2 FAQ entries for the same underlying problem.

2007/7/7, George N. White III :

...
I (slightly) abused your build.sh configuration by compiling with -std=c99 on Fedora Release 7 (i686, gcc (GCC) 4.1.2 20070502 (Red Hat 4.1.2-12)) and using system libraries (fewer chances for the compile to fail and less time to compile):

Not that it should make a difference, but does the standard build.sh work?

I have no problems with OpenSUSE 10.1 (gcc (GCC) 4.1.0 (SUSE Linux)) and 10.2 (gcc (GCC) 4.1.2 20061115 (prerelease) (SUSE Linux)) and -std=c99 -Wall.

debian stable with "-std=c99 -march=i686 -Wall" gives the same error as Fedora. Either "-std=g99 .." or "-std=c99 -D_BSD_SOURCE" works. Following the definition of strsep thru the header files and glibc sources is like those games where you get lost in a maze of twisty-turny passages. Google found some cases where people gave up trying to get configure to find strsep reliably and just coded around it. Meanwhile, SGI Irix 6.5 complains about: cc-1282 CC: ERROR File = ../../../../src/libs/xpdf/goo/GString.cc, Line = 531 More than one instance of overloaded function "pow" matches the argument list. Function symbol function "std::pow(long double, int)" is ambiguous by inheritance. Function symbol function "std::pow(double, int)" is ambiguous by inheritance. Function symbol function "std::pow(float, int)" is ambiguous by inheritance. The argument types are: (int, int). x = floor(x * pow(10, prec) + 0.5); ^ 1 error detected in the compilation of "../../../../src/libs/xpdf/goo/GString.cc". gmake[1]: *** [GString.o] Error 2 I changed this to an explicit cast to double. The result seems to work (builds formats, processes story.tex and sample2e.tex). -- George N. White III Head of St. Margarets Bay, Nova Scotia

Reply

Sign in to reply online Use email software

w.m.l＠gmx.net

8 Jul 8 Jul

5:38 p.m.

On 07.07.2007 15:14, Martin Schröder wrote:

Please test and tell me what I've forgotten. :-)

The tiny patch to #712 is missing. And what about #747? As far as I am concerned, there's nothing to complain about it. Cheers, Robert

Reply

Sign in to reply online Use email software

6381

Age (days ago)

6384

Last active (days ago)

Download

5 comments

3 participants

tags

participants (3)

George N. White III
Martin Schröder
w.m.l＠gmx.net