Subject: Re: new wv in abi cvs
From: Paul Rohr (paul@abisource.com)
Date: Tue Dec 07 1999 - 18:36:10 CST
At 03:37 PM 12/3/99 -0800, I wrote:
>Unfortunately, it looks like some leaks crept in as part of this change.
>From a quick scan of the code, it looks like this is because the logic that
>used to be in:
>
> wvReleasePAPX_FKP()
> wvReleaseCHPX_FKP()
>
>has been completely relocated to:
>
> internal_wvReleasePAPX_FKP()
> internal_wvReleaseCHPX_FKP()
>
>Presumably, the leaks happen because *sometimes* the main function also
>needs to clean stuff up. However, the obvious trivial fix -- making the
>external function always call the internal one -- just causes other
>problems.
>
>If that's not enough information to finish tracking down the leaks, let me
>know.
Here are the leaks when opening wv/examples/italic.doc:
-- snip --
Detected memory leaks!
Dumping objects ->
chp.c(656) : {3690} normal block at 0x00962120, 3 bytes long.
Data: <6 > 36 08 81
The thread 0x91 has exited with code 0 (0x0).
fkp.c(267) : {3689} normal block at 0x00961640, 24 bytes long.
Data: < ! > 00 00 00 CD 00 00 00 00 00 00 03 CD 20 21 96 00
fkp.c(266) : {3688} normal block at 0x009611F0, 3 bytes long.
Data: < > 00 FD 00
fkp.c(265) : {3687} normal block at 0x00961990, 16 bytes long.
Data: < > 00 04 00 00 07 04 00 00 0D 04 00 00 13 04 00 00
fkp.c(73) : {3676} normal block at 0x009615D0, 8 bytes long.
Data: < > 02 00 00 00 00 00 00 00
fkp.c(72) : {3675} normal block at 0x00961590, 16 bytes long.
Data: < > FD CD CD CD 00 00 00 00 00 00 00 00 00 00 00 00
fkp.c(71) : {3674} normal block at 0x00961560, 8 bytes long.
Data: < > 00 04 00 00 13 04 00 00
Object dump complete.
-- snip --
I tried to pick the smallest document I could find. More complex documents
leak even more, but they all seem related to the problem as described above.
Paul
This archive was generated by hypermail 2b25 : Tue Dec 07 1999 - 18:31:08 CST