#593 closed defect (invalid)
Memory Issues
Reported by: | Owned by: | dmorissette | |
---|---|---|---|
Priority: | high | Milestone: | |
Component: | OGR Support | Version: | 4.1 |
Severity: | normal | Keywords: | |
Cc: |
Description
I have some rather large shp files ~ 250MB (it contains river information) and when it renders even at small scales the amount of memory consumed seems a little excessive - upwards of 100MB before the memory is then released when the GGI process is exited. I have also compiled the code into a dll to implement some further functionality so the memory is not freed up completely... Seems to be happening mainly around the msOGRLayerGetAutoStyle in mapogr.c - poStyleManager->GetPart - only a small leak but this is called many times -> a blow out. Couple of possible others - if( !(p->proj = pj_init(numargs, args)) ) { in mapfile.c GDALAllRegister() in mapgdal.c
Change History (4)
comment:2 by , 20 years ago
Cc: | added |
---|---|
Owner: | changed from | to
I'll take this one. A few questions for Dave: Problem #1: You wrote that you use shapefiles, but found a leak inside msOGRLayerGetAutoStyle(). This suggests that you use STYLEITEM AUTO which is supported only by MapInfo and DGN files at the moment. So are you really using shapefiles or are you using another format? Problem #2: I'll investigate this pj_init call in mapfile.c Problem #3: GDALRegister should allocate memory only once and this memory should be freed by msCleanup() at the end of the CGI execution, if that's not the case then please provide details. If you built your own custom program then you may have to call msCleanup() at some point.
comment:3 by , 20 years ago
Resolution: | → invalid |
---|---|
Status: | new → closed |
I have run Valgrind on MapServer 4.1 CVS with: - A mapfile that uses STYLEITEM AUTO with TAB files - A mapfile that uses shapefiles with a bunch of projections objects Unfortunately (or fortunately) I get no leaks in the areas that you refer to in problems #1 and #2. There are actually no significant leaks at all with the tests that I made. This doesn't mean that there are no leaks, but just that we probably need a very specific mapfile/dataset to produce them. I'll mark the bug INVALID. Please reopen if you can provide a testcase to reproduce. We would also need to know the version of Mapserver that you're using (please test with the latest 4.1 CVS if you can), and detailed steps to reproduce.
comment:4 by , 20 years ago
Apologies - this was a bit half baked... Yes I was using .MAP files which were causing the issues - will retest and put forward a test case. I have a feeling that something to do with my configuration of the build (as a Visual C++ dll) is causing the call to delete the StylePart is not freeing the memory properly - anyhoo thanks for the prompt feedback if anyone reads this
Note:
See TracTickets
for help on using tickets.