AGG:incorrect lookup of symbol in symbolset

[cph]

when using dynamic symbols there is a lookup in the symbolset but the symbol should be looked up in the mapfile.

This problem only happens with the AGG driver.

[tbonfort]

please be more specific (layer type, symbol definition, style definition, etc) I cannot reproduce this here

[cph]

I use PCL to generate the mapserver object hierarchy. The attached works for the GD driver but not for the AGG driver. Note that if you change the line 341 from "SYMBOL 1" to "SYMBOL 0" it does not crash.

[cph]

the file which shows an AGG crash

[sdlime]

You mean line 104?

[tbonfort]

am I missing something or does symbol 1 not exist?
why are you talking about inline symbols? there are no inline symbol definitions here
"works with gd": what kind of output do you get
"crash with agg": segmentation fault, or wrong style applied, or nothing drawn ?

[cph]

sdlime: sorry yes 104

[cph]

tbonfort: as I mentioned I use mapserver only thru PCL, not generally using a mapfile definition directly. Symbol 1, does indeed not exist in the mapfile provided, but this is because that symbol is defined by PCL and added to the mapfile object.

When I change line 104 from "0" to "1", the AGG driver does not cause mapserver to segfault. When the value is "0" there is a segmentation fault in mapagg.cpp at line : 1414

[tbonfort]

ok... found it. GD just skips drawing when symbol>numsymbols will fix in next commit

[tbonfort]

fixed in r6569
thanks

[cph]

When I have "SYMBOL 1" I still get a core dump in mapagg.cpp line 1418

1417:    symbolObj *symbol = symbolset->symbol[style->symbol];;
1418:    if(!MS_VALID_COLOR(style->color) && MS_VALID_COLOR(style->outlinecolor) && symbol->type != MS_SYMBOL_PIXMAP) 

as symbol is NULL

[tbonfort]

steve:

the test posing problem is

if(style->symbol > symbolset->numsymbols || style->symbol < 0) return; /* no such symbol, 0 is OK   */

This is code that isn't specific to the agg renderer, so I wonder why there's no segfault with gd. changing the > to >= should fix this. I think the gd code should be updated too, but waiting on your comments on this as this is puzzling me.

[sdlime]

You are correct. Amazing... I suspect it hasn't come up for a couple of reasons: 1) people use symbol names, 2) you get the crash, fix the mistake and move on. Only a situation like this one would cause the error. Anyway, I've fixed in both AGG and GD code. (r6586)

Steve

[sdlime]

Closing...