[gs-devel] 8.64: issue with -dPDFA and an old compiler

[Ken Sharp]

At 08:41 23/04/2009 +0200, Giulio Orsero wrote:

>On RHEL3/5 I could create PDF/A fine, while on RH6x I would get corrupted
>PDF/A (normal PDFs were OK, only -dPDFA triggered the problem): no apparent
>errors on gs side, but Acrobat would display incomplete PDFs.
>
>I noticed that PDF/A created on the rh6x system contained strings like
>
>     /Subtype/Type1C
>     /Metadata -2147483648 0 R/Length 1834>>stream
>
>(ie 32bit overflow) while PDF/A created on RHEL3/5 had
>
>     /Subtype/Type1C/Length 1834>>stream

That was a temporary bug, it shouldn't happen with current code.


>by reading the source I saw that font metadata has been disabled so that
>"/Metadata" should never show up.

In fact that code has been removed altogether, it was only present, 
disabled, while I ran some more tests.


>Maybe you could always initialize it explicitly to cater for old/broken
>compilers?

There isn't anything strictly wrong with the compiler, its an uninitialised 
variable, so what you get is effectively random.

I'd suggest that instead you use the current code which does not emit font 
metadata. The metadata does not seem to be required (the spec merely says 
'should' contain, not 'must' contain), Acrobat never emits it and I cannot 
persuade it to validate any file containing font Metadata. Its presence 
also increases the size of the PDF file.


                     Ken