The attached file has misplaced parenthesis and the vertical text is too long (according to the customer). Distiller does a much better job. Note that 8.14 and 8.15 produce better output than HEAD. In HEAD the parenthesis on the first column are moved several pixel left, which doesn't happen on 8.1X.
Created attachment 1089 [details] FXAR3725.ps
Comparing with CPSI, paranthesis (except ones near the text "T07") are shifted down in half their width, and all other text is shifted right at half its width. Also converting to PDF with Distiller and selecting the text with Acrobat, very suspect that parenthesis are written with a rotated horizontal font (except ones near the text "T07"), but other text is with a vertical font. If so, both shifts are half pixel width.
The document depends on "1183615869 internaldict /stdCDevProc get", which Ghostscript doesn't provide, but CPSI and Distiller do provide an operator. Likely we should guess it.
Created attachment 1130 [details] cpsi.log.txt an info from CPSI.
The document supplies a different CDevProc, when stdCDevProc absents. It befhaves differently : PS>1 2 3 4 5 6 7 8 9 10 11 { pop pop pop pop pop 0 -1000 7 index 2 div 880 } exec pstack 880 0.5 -1000 0 6 5 4 3 2 1
Created attachment 1131 [details] stdCDevProc.ps We tried to define stydCDevProc to internaldict in gs_type1.ps (see attachment), but it doesn't help. Then we found that the document defines a CDevProc with each embedded font, and that procedure is called with both GS and CPSI.
When we printed the 1st ag\rgument of CDevProc (which is w0x), with GS got 0, with CPSI got many 500 and few 250. This is an incompatibility.
Patch to HEAD : http://ghostscript.com/pipermail/gs-cvs/2004-December/005137.html Patch to GS_8_1X : http://ghostscript.com/pipermail/gs-cvs/2004-December/005138.html