This is the mail archive of the cygwin mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: cygwin-gcc-fopen bug?

Lester Ingber wrote:


Hi.   On Solaris9 I use leak and there are no leaks, and I see no errors
at all.

Valgrind implements an x86 interpreter and, among other things, checks use of uninitialized memory or referencing outside of allocated memory. Dave's conjecture was that the problem was of that type, I think, rather than a leak.

One of the errors on Cywin, using the most Current Cygwin1.dll occurs
before any memory is created, ner the top of main():
  if ((Portfolio->ptr_err = fopen (infoFile, "a")) == NULL) {

I believe you are using some C++ which may do lots of things in static constructors called before main().

Now, I do have a large set of structs (mostly full of pointers for
memory to be created later in the program), e.g., several enum's with
about 40 entries, and several nested structs with about 225 entries,
in an include file common to about 50 .c files.  Could this be too
many for the Cygwin gcc default?

BTW, it is "NULL" that mangles the filename, not "NUL", likely
because I'm using sprintf() to string together paths and files.

Seems unlikely to be a capacity limit, but I'm not expert in that area. But a complex structure like that is just the sort of thing that is likely to have an initialization bug of some sort.

Your original post mentioned that your code would work with
-mno-cygwin.  If it will no longer work that way, it is
indicative that the problem is not a result of the cygwin


Unsubscribe info:
Problem reports:

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]