Add missing check for error returns from read(). Found by inspection
while reviewing the changes suggested by RVP in PR install/56303, but
not believed to be the cause of the failure reported in that PR.

Test errno when the return value from wait() indicates an error, not
when it indicates success. PR install/56303.

progress: handle EINTR in writes. PR/55914

Whitespace nit.

branches: 1.21.18;
Retry read() on EINTR. Fixes premature exit of
</dev/random progress -e cat >/dev/null

branches: 1.20.10;
Clean up #includes in progress(1).

Sort them, omit needless ones, and add omitted needful ones.

Omissions revealed by pilfering the code and trying to build it on
Mac OS X. We ought to have a better way to reveal these omissions...

branches: 1.19.2;
Use __dead

Fix argument handling of the "-b" option on 64bit platforms. Using
"-b 1024k" would previously fail with this error message:

progress: buffer size 1048576 is greater than -1.

branches: 1.17.4;
Cosmetic fix: don't exit without completing the progress bar, either on
write error or by receiving SIGPIPE. This avoids leaving the tty in a
mess. Probably addresses PR 30287.

branches: 1.16.2;
Convert to new 2 clause license

branches: 1.15.10;
Dynamically allocate the buffer used for reading data from the input
handle and default to 64k instead of the 1k (BUFSIZ) static buffer.
This makes a large difference in performance of some applications.
Make the buffer size tunable from the command line.

#include inttypes.h, for strtoimax(3).
Suggested by Slava Semushin <slava.semushin@gmail.com> in private email.

* Remove unneeded ctype.h
* remove unneeded {}s
* add some whitespace for readability

Contributed by Slava Semushin <slava.semushin@gmail.com> in private mail.

Don't try to determining the size of what's coming when reading from a pipe.
Instead of reporting some bogus values, print no progress at all.
(It's beyond me why stat(2) on a pipe returns funny values...
see the PR below for some examples!)

Fixes PR bin/33207 by myself.

OK'd by mrg@

Apply patch from PR bin/28717
This adds a -e option to progress which causes the progress bar to be
sent to stderr. This allows using progress mid-pipe, such as:
tar -cf . | progress -e bzip -1c | ssh host "cat > file"

Exit with any failure code from the cmd or gzip.
Decode gzip -l output with strtoimax() (not sscanf()) so we don't use an
uninitialised variable if the output format isn't what we expect.
This might fix some installation PRs (no error being reported)

Support strsuftoll(3) suffixes for '-l length'.
Provide an example of copying raw disks using dd, with a progress bar.

Allow custom text printed before (left of) the progress bar from progress(1):

miyu# cat openoffice-linux-1.1.0.tgz | progress -z -p 'Bytes written: ' dd of=/dev/null bs=1m
Bytes written: 193 MB 13.83 MB/s 0+195211 records in
and:
miyu# progress -f openoffice-linux-1.1.0.tgz -z -p 'Bytes written: ' dd of=/dev/null bs=1m
Bytes written: 28% |****** | 57919 KB 14.12 MB/s 00:09 ETA

OK'd by lukem.

Sigh, use an even more elaborate wait loop; it turns out we have a child
before we even start when run from sysinst. Thanks also Takao Shinohara.

Fix a race that can abort a sysinst run.
Closes bin/20275, thank you Takao Shinohara: good analysis.

This program appeared less than 3 weeks ago and it has already been
pulled up to 1.6.1? At the last minute? And with serious bugs?
ISTM that the release branch isn't supposed to work that way. It
isn't stable, it had 4 patches in 2 days, and has averaged a
patch every 4 days during its short lifetime.

branches: 1.5.2;
Make this build on non-i386 platforms.

Drop gratuitous warning for failure to get terminal width; it makes piping
progress' output ugly.

not all windows have 80 columns.

Advance a pointer correctly.

progress(1) is a standalone version of lukemftp's progress bar
(pulled in via reachover makefile) suitable for measuring the
input to arbitrary pipes.
This is intended for use by sysinst.

Test errno when the return value from wait() indicates an error, not
when it indicates success. PR install/56303.

progress: handle EINTR in writes. PR/55914

Whitespace nit.

branches: 1.21.18;
Retry read() on EINTR. Fixes premature exit of
</dev/random progress -e cat >/dev/null

branches: 1.20.10;
Clean up #includes in progress(1).

Sort them, omit needless ones, and add omitted needful ones.

Omissions revealed by pilfering the code and trying to build it on
Mac OS X. We ought to have a better way to reveal these omissions...

branches: 1.19.2;
Use __dead

Fix argument handling of the "-b" option on 64bit platforms. Using
"-b 1024k" would previously fail with this error message:

progress: buffer size 1048576 is greater than -1.

branches: 1.17.4;
Cosmetic fix: don't exit without completing the progress bar, either on
write error or by receiving SIGPIPE. This avoids leaving the tty in a
mess. Probably addresses PR 30287.

branches: 1.16.2;
Convert to new 2 clause license

branches: 1.15.10;
Dynamically allocate the buffer used for reading data from the input
handle and default to 64k instead of the 1k (BUFSIZ) static buffer.
This makes a large difference in performance of some applications.
Make the buffer size tunable from the command line.

#include inttypes.h, for strtoimax(3).
Suggested by Slava Semushin <slava.semushin@gmail.com> in private email.

* Remove unneeded ctype.h
* remove unneeded {}s
* add some whitespace for readability

Contributed by Slava Semushin <slava.semushin@gmail.com> in private mail.

Don't try to determining the size of what's coming when reading from a pipe.
Instead of reporting some bogus values, print no progress at all.
(It's beyond me why stat(2) on a pipe returns funny values...
see the PR below for some examples!)

Fixes PR bin/33207 by myself.

OK'd by mrg@

Apply patch from PR bin/28717
This adds a -e option to progress which causes the progress bar to be
sent to stderr. This allows using progress mid-pipe, such as:
tar -cf . | progress -e bzip -1c | ssh host "cat > file"

Exit with any failure code from the cmd or gzip.
Decode gzip -l output with strtoimax() (not sscanf()) so we don't use an
uninitialised variable if the output format isn't what we expect.
This might fix some installation PRs (no error being reported)

Support strsuftoll(3) suffixes for '-l length'.
Provide an example of copying raw disks using dd, with a progress bar.

Allow custom text printed before (left of) the progress bar from progress(1):

miyu# cat openoffice-linux-1.1.0.tgz | progress -z -p 'Bytes written: ' dd of=/dev/null bs=1m
Bytes written: 193 MB 13.83 MB/s 0+195211 records in
and:
miyu# progress -f openoffice-linux-1.1.0.tgz -z -p 'Bytes written: ' dd of=/dev/null bs=1m
Bytes written: 28% |****** | 57919 KB 14.12 MB/s 00:09 ETA

OK'd by lukem.

Sigh, use an even more elaborate wait loop; it turns out we have a child
before we even start when run from sysinst. Thanks also Takao Shinohara.

Fix a race that can abort a sysinst run.
Closes bin/20275, thank you Takao Shinohara: good analysis.

This program appeared less than 3 weeks ago and it has already been
pulled up to 1.6.1? At the last minute? And with serious bugs?
ISTM that the release branch isn't supposed to work that way. It
isn't stable, it had 4 patches in 2 days, and has averaged a
patch every 4 days during its short lifetime.

branches: 1.5.2;
Make this build on non-i386 platforms.

Drop gratuitous warning for failure to get terminal width; it makes piping
progress' output ugly.

not all windows have 80 columns.

Advance a pointer correctly.

progress(1) is a standalone version of lukemftp's progress bar
(pulled in via reachover makefile) suitable for measuring the
input to arbitrary pipes.
This is intended for use by sysinst.