crashpad_database_util: Accept --new-report=- to read a new report from standard input

crashpad_database_util: Accept --new-report=- to read a new report from
standard input.

R=rsesek@chromium.org

Committed: https://chromium.googlesource.com/crashpad/crashpad/+/124ace19bdce69e2974e3cfe3ef166241c5be8ec

[Mark Mentovai, 2015-03-20 21:38:16]

mark@chromium.org changed reviewers:
+ rsesek@chromium.org, scottmg@chromium.org

[Mark Mentovai, 2015-03-20 21:43:33]

This enhances and depends on https://codereview.chromium.org/1018853006/.

Robert: basic review, since you reviewed the rest of crashpad_database_util.

Scott: sanity-check what I did for Windows here. Even though
crashpad_database_util doesn’t currently work on Windows, I added support for
the stdio part in util. However, it occurs to me that we wouldn’t ever actually
want to use the CRT stdio wrapper around stdin on Windows, just like we
typically use Win32 directly elsewhere. We might just prefer to use
GetStdHandle(STD_INPUT_HANDLE) with normal HANDLE-based I/O. If the CRT stdio
family would never be the ideal way to deal with stdin on Windows, I’ll just #if
it to be OS_POSIX only.

[scottmg, 2015-03-20 22:00:19]

On 2015/03/20 21:43:33, Mark Mentovai wrote:
> This enhances and depends on https://codereview.chromium.org/1018853006/.
> 
> Robert: basic review, since you reviewed the rest of crashpad_database_util.
> 
> Scott: sanity-check what I did for Windows here. Even though
> crashpad_database_util doesn’t currently work on Windows, I added support for
> the stdio part in util. However, it occurs to me that we wouldn’t ever
actually
> want to use the CRT stdio wrapper around stdin on Windows, just like we
> typically use Win32 directly elsewhere. We might just prefer to use
> GetStdHandle(STD_INPUT_HANDLE) with normal HANDLE-based I/O. If the CRT stdio
> family would never be the ideal way to deal with stdin on Windows, I’ll just
#if
> it to be OS_POSIX only.

Generally stdin/- isn't very useful on Windows because there's too many way to
get munged by CRLF translation (we'd probably at least need to
`_setmode(_fileno(stdin), _O_BINARY)`). But I think it's fine not to clutter
with #ifs for now so that it at least compiles on Windows, and we can decide
later if it's useful once the supporting code works.

[Mark Mentovai, 2015-04-17 21:33:52]

Robert, this one’s still hanging out too.

[Robert Sesek, 2015-04-23 13:40:13]

https://codereview.chromium.org/1023943003/diff/20001/util/file/file_reader.h
File util/file/file_reader.h (right):

https://codereview.chromium.org/1023943003/diff/20001/util/file/file_reader.h...
util/file/file_reader.h:157: class WeakStdioFileReader : public
FileReaderInterface {
Why use this instead of WeakFileHandleReader(STDIN_FILENO) ?

[Mark Mentovai, 2015-08-06 14:30:08]

https://codereview.chromium.org/1023943003/diff/20001/util/file/file_reader.h
File util/file/file_reader.h (right):

https://codereview.chromium.org/1023943003/diff/20001/util/file/file_reader.h...
util/file/file_reader.h:157: class WeakStdioFileReader : public
FileReaderInterface {
Robert Sesek wrote:
> Why use this instead of WeakFileHandleReader(STDIN_FILENO) ? 

I did try WeakFileHandleReader(fileno(stdin)) first, but it didn’t work as well.
It totally misses the EOF when stdin is a terminal:

mark@mela zsh% out/Debug/crashpad_database_util -d /tmp/z --new-report=-  
crash report
^D
[…wait…]

It only hears it when you press ^D twice in a row. Weird. The loop in ReadFile()
gets the whole string on the first attempt, but then goes back to block even
though the ^D should have closed it and the next read should immediately return
0.

[Mark Mentovai, 2015-08-06 14:47:15]

Patchset #4 (id:60001) has been deleted

[Mark Mentovai, 2015-08-06 14:49:39]

Never mind, I figured out why it was hanging with WeakFileHandleFileReader and
fixed it. Updated.

[Mark Mentovai, 2015-08-06 21:01:08]

I take it back. WeakFileHandleFileReader doesn’t work because fileno(stdin) and
STDIN_FILENO aren’t valid file handles on Windows.

Updated again.

[Robert Sesek, 2015-08-07 16:21:02]

LGTM

[Mark Mentovai, 2015-08-07 17:57:10]

Committed patchset #6 (id:120001) manually as
124ace19bdce69e2974e3cfe3ef166241c5be8ec (presubmit successful).