Opened 5 years ago

Closed 3 years ago

Last modified 3 years ago

#57595 closed defect (invalid)

tiff broken by zstd

Reported by: gaming-hacker (G Alexander) Owned by: ryandesign (Ryan Carsten Schmidt)
Priority: Normal Milestone:
Component: ports Version: 2.5.4
Keywords: Cc: seanfarley (Sean Farley)
Port: tiff zstd

Description (last modified by mf2k (Frank Schima))

tiff portfile was hacked to remove webp and jbigkit; this is unnecessary because zstd portfile is incorrect as it is wrong.

zstd bombs because this

build.target        allmost

should be

build.target        default

i build tiff with all options without error

Change History (5)

comment:1 Changed 5 years ago by mf2k (Frank Schima)

In the future, please use WikiFormatting.

comment:2 Changed 5 years ago by mf2k (Frank Schima)

Cc: ryandesign removed
Description: modified (diff)
Owner: set to ryandesign
Status: newassigned

comment:3 in reply to:  description Changed 5 years ago by ryandesign (Ryan Carsten Schmidt)

Port: zstd added

In what way is "tiff broken by zstd"? It worked for me on High Sierra before I committed it. Could you attach a main.log showing what problem you experienced?

I initially considered making tiff's zstd support a variant, since it's new so probably nothing requires it yet. I decided against that, because we prefer there to be fewer variants, but if tiff's zstd support is causing build failures that we can't otherwise resolve, I can still change it to be a variant.

I did see after committing it that zstd 1.3.4 failed to build for Snow Leopard i386 (whether building for i386 alone or building i386 x86_64 universal). I would like to fix that but haven't yet looked into it.

I see that zstd 1.3.7 is available. The port should be updated. Maybe it fixes some of these problems.

Replying to gaming-hacker:

tiff portfile was hacked to remove webp

Not a hack particularly, just routine iterative improvements to MacPorts. In updating tiff to 4.0.10, I saw that it had new support for webp and zstd so I enabled them. Then I realized after committing, by examining build failures on the buildbot, that I had introduced a circular dependency: tiff depended on webp, but webp already depended on tiff. So I removed the webp support from tiff to fix that.

In fact, the webp library does not require tiff. It's only the webp programs that do. But at present both the webp programs and the webp libraries are installed by the same webp port. I intend to reenable webp support in tiff in the future by splitting the webp library out into a new libwebp subport. But that will take a lot more effort, and I needed to commit a fix immediately to prevent a zillion build failures, because tiff is a very common port, depended on by a lot of other ports.

and jbigkit;

jbigkit isn't removed; it's there as a variant, as it has been for over 8 years. I briefly looked into enabling it by default, but noticed that jbigkit is licensed GPL-2+ while tiff is licensed BSD. I suspected that enabling jbigkit by default might make some ports that depend on jbigkit nondistributable. I didn't look into it very closely and erred on the side of not changing it at this time.

this is unnecessary because zstd portfile is incorrect as it is wrong.

zstd bombs because this

build.target        allmost

should be

build.target        default

i build tiff with all options without error

In what way is build.target default better than build.target allmost? (It may be, but if you could explain your reasoning that may convince Sean to make the change.)

comment:4 Changed 3 years ago by ryandesign (Ryan Carsten Schmidt)

Resolution: invalid
Status: assignedclosed

Closing since we didn't get clarification about what the perceived problem was.

comment:5 Changed 3 years ago by gaming-hacker (G Alexander)

Sorry forgot about this.

Note: See TracTickets for help on using tickets.