What they said. For fieldwork, you get a flat-file database for organising your field notes and automatically generating glosses and dictionaries. (A relational database is overkill.) Toolbox (The Field Linguist’s Toolbox) and its predecessor The Linguist’s Shoebox from SIL International are the default tools.
Databases are less useful than you might think, though I found they were useful for typological work (if you’re doing wide surveys of languages).
For Computational Linguistics, you learn Perl (if you’re living 20 years ago like I still am) or Python, and you get hold of a good software library. The premier one seems to be Natural Language Toolkit. Treat it as a starting point, but it is a very good starting point. Enough that I’ve winced and taught myself enough Python to use it, though I still find Python distasteful.
If you’re doing phonetics, you will be doing a lot of IT, to get the data, and to get statistics about the data: phonetics is a lot closer to disciplines like psychology. Like Joonas Vakkilainen, you will get familiar with R. (Or Python, I guess.) Ditto sociolinguistics, as Joonas said. But for most other fields of linguistics, you won’t likely need more stats than you can get out of Excel.
Sorry for 4 month delay in A2A, Z-Kat.