Em qui., 17 de out. de 2024 às 07:30, Alexander Lakhin <exclus...@gmail.com> escreveu:
> Hello Ranier, > > 16.10.2024 14:14, Ranier Vilela wrote: > > Em qua., 16 de out. de 2024 às 04:45, Peter Eisentraut < > pe...@eisentraut.org> escreveu: > >> >> This wouldn't fix anything, I think. If colors is NULL, then strsep() >> already returns NULL, so the added code does nothing. >> > If *colors* is NULL, then the delimiter is not found and strsep will > return the entire > string **stringp, so the token becomes invalid*. > > IMO, I think it must be necessary to check if *colors* are NULL too. > > > I've tested your proposed change and what I'm seeing is that: > PG_COLOR=always PG_COLORS="error=01;31" initdb > doesn't color the "error" word: > > while with only Peter's patch it works as expected: > > > Does your change work differently for you? > No. Thanks for the test. It seems to me that with strsep, the only alternative is to run the risk of processing an invalid token? I ran the test with Peter's patch in Windows and the terminal doesn't color the "error" word, perhaps this feature does not work with Windows. [image: error1.png] I withdraw the proposed patch. best regards, Ranier Vilela
