Apache OpenOffice (AOO) Bugzilla – Full Text Issue Listing |
Summary: | [From Symphony] There is a memory leak in function SvBaseLink::~SvBaseLink() | ||||||
---|---|---|---|---|---|---|---|
Product: | performance | Reporter: | ChaoHuang <chao.dev.h> | ||||
Component: | code | Assignee: | AOO issues mailing list <issues> | ||||
Status: | CLOSED FIXED | QA Contact: | |||||
Severity: | Normal | ||||||
Priority: | P3 | CC: | Armin.Le.Grand, liushenf | ||||
Version: | AOO 3.4.0 | ||||||
Target Milestone: | 4.0.0 | ||||||
Hardware: | All | ||||||
OS: | All | ||||||
Issue Type: | DEFECT | Latest Confirmation in: | --- | ||||
Developer Difficulty: | --- | ||||||
Issue Depends on: | |||||||
Issue Blocks: | 120975, 121366 | ||||||
Attachments: |
|
Description
ChaoHuang
2012-06-20 07:07:57 UTC
The member data pImpl typed BaseLink_Impl is created in contructor. But it does not released in destructor. So it is a memory leak. Created attachment 78401 [details]
for file "main/sfx2/source/appl/lnkbase2.cxx"
ALG: SvBaseLink has three constructors, and only in two of them pImpl gets allocated and thus initialized. Thus, in the 3rd constructor, pImpl will point to a random address. Freeing it in the destructor could crash the office. A minimal fix would have to init pImpl to zero and check before construction. But even then, the implementation makes use of pImpl without tests, so maybe in the 3rd constructor it should also be allocated. All in all, the whole class would need more rework, all members initialized in all constructors, and pImpl decided for 3rd constructor. ALG: Added C++ style initialisations to all constructors, secured destruction of pImpl, secured all accesses to pImpl if not initialized, asserting if not. Commited as r1352129. (In reply to comment #3) > ALG: SvBaseLink has three constructors, and only in two of them pImpl gets > allocated and thus initialized. Thus, in the 3rd constructor, pImpl will > point to a random address. Freeing it in the destructor could crash the > office. > > A minimal fix would have to init pImpl to zero and check before > construction. But even then, the implementation makes use of pImpl without > tests, so maybe in the 3rd constructor it should also be allocated. All in > all, the whole class would need more rework, all members initialized in all > constructors, and pImpl decided for 3rd constructor. From my point of view, pImpl is a fundamental member in class SvBaseLink, which should be initialized in every constructor. There is no interface to set the value for pImpl. What's more, there is no checking for pImpl before use it, and pImpl is initialized in the first two constructors. I think that it should be also initialized in the third constructor. Another concern I want to say is that it should be well to add an assert in the beginning of the function than in the end. I prefer code style ------------------------------------------------------------- void SvBaseLink::SetLinkManager( LinkManager* _pMgr ) { OSL_ENSURE(pImpl!=NULL, "No pImpl (!)"); pImpl->m_pLinkMgr = _pMgr; } ------------------------------------------------------------- to ------------------------------------------------------------- void SvBaseLink::SetLinkManager( LinkManager* _pMgr ) { if(pImpl) { pImpl->m_pLinkMgr = _pMgr; } else { OSL_ENSURE(false, "No pImpl (!)"); } } ------------------------------------------------------------- Suggest to put it into AOO 3.5.0 release |