From: Peter P. <pet...@gm...> - 2011-02-25 20:14:12
|
Hi Egon, Thanks for your fast reply. On Friday 25 February 2011 20:28:38 Egon Willighagen wrote: [...] > I am not an active Strigi developer, but wrote strigi plugins in the > past for chemistry file. File formats are difficult to parse, in > chemistry, and surely in other domains too. The problem is that there > is much software around that does not follow standards, which makes > extracting information a hard problem. I agree that parsing is not easy and it is sometimes very tricky to consider all the strange things that are part of some file formats. I wrote several parsers some years ago when working for another company... But I'm really convinced that writing parsers in a way that they at least don't crash when getting invalid/corrupt/strange files as input is not that hard. > As such, the list is really not that bad in my opinion, and as a Java > developers it would not have mattered much, as you would not have > gotten segfault for this... I wonder if properly catching these > crashes in C++ is really not possible at all... > > But, that's not what I want to reply about. As said, input is crufty > and you cannot compensate for all. Compensating is not a must-have, but at least not crashing is really a must and not that difficult (at least when not doing crazy unchecked pointer-arithmetics like I've seen in a few analyzers...) > That must never cause segfaults, of > course. Now fixing these issues is tricky, as these segfaults are > relatively rare. I am not good at reading C++ stacktraces, and note a > lot of threading, complicating the output... > > Some bug reports do, in fact, ask for the files that cause the test; > IMHO that is crucial here, and should be part of the bug report. Some > bug reporters actually do report that, and that is very useful. Yes. > Now, a simple test to way to reproduce these crashes and get a cleaner > stacktrace would be to use the xmlindexer instead of dolphin, and > that's what I like to provide as feedback here. Please ask the bug > reporters that can reproduce the bug and know the file, to report what > xmlindexer does on those files. I fully agree, but we did this already. We attached files to the bug-reports or referred from the strigi bug tracker to the corresponding kde bug report where the file is attached. Still no reply for many, many reports... (not all of course, there has also been support from a few people). So the reporting does not seem to be the issue - it is fine for me if the maintainer of an analyzer tells me that he is currently busy and cannot fix this bug during the next months. But getting no answer at all for so many issues is quite concerning... :-( > The least that would do, if give stacktraces without threading going around. > > Egon > > > |