BUG: read_csv with bad file coreing

[jreback]

from ML:
https://groups.google.com/forum/#!topic/pydata/KO-PmQdBUZI

In [1]: data = """c1\ntext11,text12\ntext21,text22"""

In [2]: read_csv(StringIO(data))
Out[2]: 
            c1
text11  text12
text21  text22

In [3]: read_csv(StringIO(data),header=0,names=list('abc'))
Bus error (core dumped)

[jreback]

cc @guyrt

[guyrt]

Confirmed that including too many names can cause a segfault:

data = """\                                  
1,2,3
4,5,6"""

df = pd.read_csv(StringIO(data), header=0, names=['a', 'b', 'c', 'd'])
Segmentation fault

Looks like a problem with header. The fix to #4335 fixes this for python parser:

df = pd.read_csv(StringIO(data), header=0, names=['a', 'b', 'c', 'd'], engine='python')
ValueError: Number of passed names did not match number of header fields in the file

Looks like we need a similar fix for C engine.

[jreback]

gr8! i'll move this back to 0.13 then

[jreback]

@guyrt I think this would be gr8 to include in 0.13; going to cut the release candidate early next week.

[guyrt]

I'll see what I can do. I've identified the problem: we overrun the parser buffer. Just working on a fix.

[jreback]

gr8 thanks!