The code is currently correct since in the only place we create a HashEntry half initialized we set the other members manually, but there doesn't seem to be a reason to not initialize all members in the constructor, and this would get rid of some coverity warnings.

I haven't looked into just what we'd need to change for this, but it would make me sad if it results in first zero-initializing the members, only to overwrite them with the real values.

Just need to initialize the members in the constructor here instead of doing it manually a few lines below: https://hg.mozilla.org/mozilla-central/file/a5af73b32ac8/gfx/thebes/gfxGlyphExtents.cpp#l133

Je souhaites m'occuper de ce bug [comme dit sur irc].

(In reply to robin.moussu+git@gmail.com from comment #3) > Je souhaites m'occuper de ce bug [comme dit sur irc]. Cool (let's do it in English on bugzilla, though).

> Cool (let's do it in English on bugzilla, though). Sure, sorry for being messy :)

So, I have a class `nsUint32HashKey`, with some public attribute > public: > float x, y, width, height; and a constructor > explicit HashEntry(KeyTypePointer aPtr) : nsUint32HashKey(aPtr) {} which leave those attributes non-initialised. The function > void gfxGlyphExtents::SetTightGlyphExtents(uint32_t aGlyphID, const gfxRect& aExtentsAppUnits) Initialize those attributes. So what I want to do is to merge the constructor and SetTightGlyphExtents() (so add the two parameters of SetTightGlyphExtents to the constructor, and remote SetTightGlyphExtents()), and set the attributes private. Am I right? NB: SetTightGlyphExtentsvd is only use in gfx/thebes/gfxFont.cpp

edit: SetTightGlyphExtents() is a function of the parent of class `nsUint32HashKey`, so of course, I do not remove that function.

You can simply add the missing members to the constructor and SetTightGlyphExtents would look something like > mThighGlyphExtents.PutEntry(HashEntry(aGlyphID, > aExtentsAppUnits.X(), > [etc.] No need to set the attributes private for now, we just want to shut coverity's uninitialized-member warnings up.

A night of sleep was a good thing. I follow a bit the use graph of HashEntry. What I understood, is that an nsTHashTable<HashEntry> is created in GfxGlyphExtents. HashEntry are half-initialized. When we call GfxGlyphExtents.SetTightGlyphExtents(), they are set, and there is obvious link between the creation of the object to SetTightGlyphExtents(). So I think that the only think I can do is first zero-initializing the members of HashEntry, only to overwrite them with the real values in SetTightGlyphExtents().

@Nicolas Silva: Do you think that zero-initializing is a good thing ?

(In reply to robin.moussu+git@gmail.com from comment #9) > A night of sleep was a good thing. > Actually, sorry, I am the one who needs a good night of sleep. I read the code backwards (I thought we could also pass an entry as parameter of PutEntry but you only pass the key) and it would be wasteful to zero-initialize the entry and set the members right after that just to get rid of a benign coverity warning so I'll close the bug and mark the warning as false-positive.