#3944 closed defect (fixed)
COMMON: readLine_NEW fails to read last line in file
| Reported by: | salty-horse | Owned by: | wjp |
|---|---|---|---|
| Priority: | normal | Component: | --Other-- |
| Version: | Keywords: | ||
| Cc: | Game: |
Description
Using latest svn r34343.
If the last "line" in a file doesn't end with a newline, readLine_NEW will return 0 instead of the real data.
In my case, I am opening a file with Common::File::open() and calling readLine_NEW on it.
Suppose the last characters of the file are 'a', 'b' followed by EOF.
readLine_NEW will execute "c = readByte();" in common/stream.cpp:166 twice. After the third time, c will be '\0'. It then gets to the check:
if (ioFailed() || (len == 0 && eos())) return 0;
ioFailed() returns True and then 0 is returned, even if len > 0.
Maybe the real problem is ioFailed returning 'true' when nothing is actually failing. It's set when readByte() calls File::read(&b, 1).
AFAIK, this doesn't actually cause any errors in existing engines. I've encountered it in my KoM engine.
Ticket imported from: #2095303. Ticket imported from: bugs/3944.
Attachments (1)
Change History (7)
by , 16 years ago
| Attachment: | readline_fix.patch added |
|---|
comment:1 by , 16 years ago
comment:2 by , 16 years ago
| Owner: | set to |
|---|
comment:3 by , 16 years ago
Fixed thanks to the recent Stream::eos() work by wjp. Confirmed by a unit test, at least for MemoryReadStream.
File streams *should* also work, if not, please reopen this item. I hope that we can add a unit test for those eventually, too, but this will require improving the "make test" linking system, as we'd have to compile in backends/ into our unit test runner.
comment:4 by , 16 years ago
| Owner: | changed from to |
|---|---|
| Resolution: | → fixed |
| Status: | new → closed |
comment:6 by , 6 years ago
| Component: | → --Other-- |
|---|
Attaching a proposed fix.
It changes File::read() to *not* set ioFailed if it's the end of the file, and adds a check for eos() in readLine_NEW() File Added: readline_fix.patch