Package: gv Version: 1:3.6.1-9 Severity: normal In certain cases the ps file scanner can get confused by ps files embedded within a page. The file <http://www.inference.phy.cam.ac.uk/sanjoy/bug/gv/several-pages.ps>, extracted from a colleague's textbook, shows the problem. Running gv several-pages.ps shows only page "540,54" (really "540,541") in the page list. But the file has a few pages: (540,541) and (543,542) and (639,638). On my laptop (Pentium 3) all versions of gs (v8.01, 8.14, and 8.50) show the full set of pages, so I the problem is most likely in gv.
The <http://www.inference.phy.cam.ac.uk/sanjoy/bug/gv/difficult-page.ps> file has the difficult page itself, without the huge prolog (all those tex fonts) or the pages after it. Also, the vanilla gv 3.6.1, say from the gv_3.6.1.orig.tar.gz file, shows all the pages in the sidebar. So the problem is probably in a Debian patch. By trying various partial patches and guessing a bit, I narrowed it down to this hunk from debian/patches/02_293976=045735_filename+unknown.diff (available after doing 'apt-src install gv/unstable'), which applies to the readline() routine: --- gv-3.6.1.orig/src/ps.c +++ gv-3.6.1/src/ps.c @@ -1826,7 +1858,12 @@ else #endif if (!IS_COMMENT("Begin")) {} /* Do nothing */ - else if IS_BEGIN("Document:") SKIP_UNTIL_1("EndDocument") + else if IS_BEGIN("Document:") { /* Skip the EPS without handling its content */ + while (line && !IS_END("Document")) { + line = ps_io_fgetchars(fd,-1); + if (line) *line_lenP += FD_LINE_LEN; + } + } else if IS_BEGIN("Feature:") SKIP_UNTIL_1("EndFeature") #ifdef USE_ACROREAD_WORKAROUND else if IS_BEGIN("File") SKIP_UNTIL_2("EndFile","EOF") (also available at <http://www.inference.phy.cam.ac.uk/sanjoy/bug/gv/causes-bug.diff>) The culprit seems to be the change from SKIP_UNTIL_1("EndDocument") to the while loop. SKIP_UNTIL_1() calls readline(), which skips over embedded ps documents. The while loop, however, blindly reads to the next %%EndDocument. The difference shows up when an included ps document includes a ps document, as the troublesome page does. Here is its structure: $ egrep '^%%(Begin|End)Document' difficult-page.ps %%BeginDocument: /home/mackay/handbook/figs/inter/sc_a1.blank.ps %%BeginDocument: EPSF %%EndDocument %%EndDocument %%BeginDocument: /home/mackay/handbook/figs/inter/sc_a2b.blank.ps %%BeginDocument: EPSF %%EndDocument %%EndDocument %%BeginDocument: /home/mackay/handbook/figs/inter/sc_a3.blank.ps %%BeginDocument: EPSF %%EndDocument %%EndDocument %%BeginDocument: /home/mackay/handbook/figs/test_train.eps %%EndDocument %%BeginDocument: /home/mackay/handbook/figs/probability.eps %%EndDocument %%BeginDocument: /home/mackay/handbook/figs/error_bar_net.ps %%EndDocument So the first document includes another document. With the patch, the scanner will get to the first BeginDocument, then skip to the first EndDocument, which matches the *second* BeginDocument. But the second BeginDocument gets skipped over, so the scanner should be quite confused! Now, I'm not sure whether the analysis above is totally right. But applying the patch above produces the bug, and reverting it removes the bug. And the behavior with the patch seems to be incorrect and worth fixing. Although perhaps the embedded ps file is part of the problem? It was created by gnuplot and is basically an EPS file surrounded by a few lines to turn it into a regular PS file -- probably one is asking for trouble by including it. That section of 'several-pages.ps' begins: %%BeginDocument: /home/mackay/handbook/figs/inter/sc_a2b.blank.ps %!PS-Adobe-2.0 %%Creator: gnuplot %%DocumentFonts: Courier %%DocumentNeededFonts: Courier %%Pages: (atend) %%BoundingBox: 46 50 550 770 %%EndComments /GnuTransform { 90 rotate 50 -550 translate 2 2 scale } def %%EndProlog %%Page: ? 1 gsave GnuTransform %%BeginDocument: EPSF 1 dict begin /showpage {} def %!PS-Adobe-2.0 EPSF-2.0 %%Creator: gnuplot %%DocumentFonts: Courier %%DocumentNeededFonts: Courier %%BoundingBox: 0 0 360 252 %%EndComments So gnuplot was doing something funky by making an EPS file and wrapping a few lines of ps around it to handle the scaling and moving the figure towards the center of the page. -- System Information: Debian Release: 3.1 APT prefers testing APT policy: (990, 'testing'), (500, 'unstable') Architecture: i386 (i686) Kernel: Linux 2.4.27-200503091 Locale: LANG=en_US, LC_CTYPE=en_US (charmap=ISO-8859-1) Versions of packages gv depends on: ii gs 8.01-5 Transitional package ii gs-afpl [gs] 8.14-3 The AFPL Ghostscript PostScript in ii gs-gpl [gs] 8.01-5 The GPL Ghostscript PostScript int ii libc6 2.3.2.ds1-20 GNU C Library: Shared libraries an ii libice6 4.3.0.dfsg.1-10 Inter-Client Exchange library ii libsm6 4.3.0.dfsg.1-10 X Window System Session Management ii libx11-6 4.3.0.dfsg.1-10 X Window System protocol client li ii libxext6 4.3.0.dfsg.1-10 X Window System miscellaneous exte ii libxmu6 4.3.0.dfsg.1-10 X Window System miscellaneous util ii libxpm4 4.3.0.dfsg.1-10 X pixmap library ii libxt6 4.3.0.dfsg.1-10 X Toolkit Intrinsics ii xaw3dg 1.5+E-8 Xaw3d widget set ii xlibs 4.3.0.dfsg.1-10 X Keyboard Extension (XKB) configu -- no debconf information -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
