Hi everyone,
Thanks to Ulrike's test files I could track it down ...
> So it seems that the $objstream is *only* initialized when the objdict[$count]
> is true, i.e., if the obj line looks like
> 3 0 obj <<
> But the << is on the next line, and even then, moving it up, did not change
> anything.
That was absolutely true, but my check was wrong, because thumbpdf was
parsing a *different* file, the thumbpdf.pdf.
There are two changes in PDF-X or whatever that adds these new lines
after << and >>.
With the attached patch it works again on my installations.
But I don't know whether
tell
seek
works on Windows, too (Ulrike ... please ...)
The idea is that we read the next line (peek at it), and if it
is a << or endobj then we just do as normal, otherwise we put
the line back.
Patch attached, comments welcome.
Thanks
Norbert
------------------------------------------------------------------------
PREINING, Norbert http://www.preining.info
JAIST, Japan TeX Live & Debian Developer
GPG: 0x860CDC13 fp: F7D8 A928 26E3 16A1 9FA0 ACF0 6CAC A448 860C DC13
------------------------------------------------------------------------
--- thumbpdf.pl.orig 2014-07-13 16:24:53.823731368 +0900
+++ thumbpdf.pl 2014-07-15 00:29:21.030833482 +0900
@@ -937,6 +937,18 @@
$objno[$count] = $1;
$getobjindex[$1] = $count;
$objdict[$count] = ($2); # boolean (if $2 exists)
+ if (!$objdict[$count]) {
+ # check for << on thext line
+ my $pos = tell();
+ $_ = <PDF>;
+ if (/^<<$/) {
+ $objdict[$count] = 1;
+ $lineno++;
+ } else {
+ # undo the reading
+ seek PDF, $pos, 0;
+ }
+ }
my $stream = 0;
print "* obj $objno[$count]" .
(($objdict[$count]) ? " (dict)" : "") .
@@ -952,7 +964,20 @@
{
if (/^>>/)
{
- last if /^>>\s+endobj$/; # obj without stream
+ # new line after >>
+ if (/^>>\s+endobj$/) {
+ last; # obj without stream
+ } else {
+ my $pos = tell();
+ $_ = <PDF> ;
+ if (/^endobj$/) {
+ last; # obj without stream
+ $lineno++;
+ } else {
+ seek PDF, $pos, 0;
+ }
+ }
+ # last if /^>>\s+endobj$/; # obj without stream
# get stream
$_ = <PDF>; $lineno++;
